[{"data":1,"prerenderedAt":1057},["ShallowReactive",2],{"blog-article-en-ai-recruitment-hr-ai-act-high-risk":3,"blog-related-en-ai-recruitment-hr-ai-act-high-risk":740},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"seoTitle":10,"locale":7,"translationSlug":11,"date":12,"updated":13,"image":14,"author":15,"tags":18,"takeaways":23,"faq":29,"body":45,"_type":734,"_id":735,"_source":736,"_file":737,"_stem":738,"_extension":739},"/blog/ai-recruitment-hr-ai-act-high-risk","blog",false,"en","AI in Recruitment and HR: What to Automate and What the AI Act Treats as High Risk","What AI can safely do in HR, what the AI Act classifies as high risk in recruitment, and how to design systems that support decisions without making them.","AI in Recruitment and HR: Where High Risk Begins","ia-recrutamento-rh-ai-act-alto-risco","2026-10-08","2026-11-14","/images/blog/ia-recrutamento-rh.jpg",{"name":16,"role":17},"Pedro Santos","Principal Software Architect",[19,20,21,22],"Artificial Intelligence","Compliance","Automation","Custom Software",[24,25,26,27,28],"In human resources, AI can comfortably take on almost all the work that surrounds a decision: scheduling interviews, replying to candidates, running onboarding, explaining internal policies and processing documents.","What the AI Act classifies as high risk is whatever touches the decision about people: filtering and ranking applications, assessing candidates, supporting promotions or dismissals and monitoring performance. An AI module switched on in your recruitment software can put the company in that category without anyone having decided it.","Following the postponement that came into force in July 2026, the high-risk obligations for these systems apply from 2 December 2027. The ban on inferring emotions in the workplace, the transparency duties, the duty to inform employees and the GDPR already apply today.","Designing AI to support rather than decide is an architecture question: criteria set by people, evidence attached to every suggestion, a human decision recorded with its reasoning, and bias testing before anything goes live.","The safest route is to start with what is low risk and frees up the most hours, usually the employee portal and onboarding, and only then decide, with real information, whether it makes sense to touch candidate screening.",[30,33,36,39,42],{"q":31,"a":32},"Is it forbidden to use AI to screen CVs?","It is not forbidden, but the AI Act classifies systems used to analyse and filter applications or to assess candidates as high risk. That brings obligations around human oversight, record keeping, data quality and informing the people affected. What has been prohibited since February 2025 are practices such as inferring the emotions of employees or candidates in the workplace.",{"q":34,"a":35},"When do the AI Act high-risk obligations apply to recruitment?","The original timeline pointed to August 2026. The simplification package known as the Digital Omnibus, published as Regulation (EU) 2026/1744 and in force since 27 July 2026, pushed back the obligations for Annex III systems, which is where employment sits, to 2 December 2027. The timeline has already been revised once, so it is worth confirming with your legal team before fixing any plans.",{"q":37,"a":38},"If a person clicks the final button, does the decision stop being automated?","Not necessarily. If the decision maker simply confirms the list the system ranked, without the time, information or authority to overrule it, the human involvement is only apparent. For it to count, that person has to see the evidence, be able to disagree and leave a record of why they decided as they did.",{"q":40,"a":41},"Is an assistant that answers employee questions about holidays and internal policies high risk?","As a rule, no, because it makes no decisions about people: it gives information drawn from company documents. It does have to make clear that it is an automated system, cite the source of every answer and hand over to a person any case that depends on a decision, such as an exception to the holiday policy.",{"q":43,"a":44},"Does the AI module in our recruitment software already take care of compliance?","It takes care of the part that falls to the vendor. The company using the system still has its own obligations: ensuring oversight by trained people, keeping records, informing candidates and employees, and complying with the GDPR. It is worth asking the vendor for the system documentation before switching it on.",{"type":46,"children":47,"toc":712},"root",[48,56,61,68,73,79,84,89,95,100,165,170,176,181,320,325,330,335,340,346,351,384,390,400,410,420,426,431,441,451,461,471,481,494,499,505,512,517,523,528,534,539,545,550,556,561,567,630,642,648,660,665,671,699],{"type":49,"tag":50,"props":51,"children":52},"element","p",{},[53],{"type":54,"value":55},"text","Picture a logistics operator with a few hundred employees, gearing up for its peak season.\nThere are twelve open roles for warehouse operatives, and in ten days the advert\nhas drawn more than six hundred applications. The HR team is two people working\nfull time, who also handle holidays, absences, occupational health certificates\nand last week's new starters. In the latest update, the recruitment software\ngained a new button: \"rank candidates by fit with AI\". One click switches it on\nand, suddenly, the six hundred applications appear sorted from 98 down to 12.",{"type":49,"tag":50,"props":57,"children":58},{},[59],{"type":54,"value":60},"The temptation is easy to understand. It is also the exact point at which the\ncompany walks, without ever having decided to, into the category the European\nregulation on artificial intelligence treats most strictly: a number attached\nto a person's name, used to decide who gets invited to interview, is precisely\nthe kind of system the AI Act set out to regulate.",{"type":49,"tag":62,"props":63,"children":65},"h2",{"id":64},"short-answer",[66],{"type":54,"value":67},"Short answer",{"type":49,"tag":50,"props":69,"children":70},{},[71],{"type":54,"value":72},"AI can safely handle almost all the work around HR decisions: scheduling\ninterviews, replying to candidates, answering questions about internal\npolicies, running onboarding and processing documents. The AI Act treats as\nhigh risk whatever touches the decision about people: filtering and ranking\napplications, assessing candidates, supporting promotions or dismissals,\nallocating tasks based on personal traits and monitoring performance. The\nhigh-risk obligations apply to these systems from 2 December 2027, but the ban\non inferring emotions in the workplace, the transparency duties and the GDPR\nalready apply today.",{"type":49,"tag":62,"props":74,"children":76},{"id":75},"ai-can-do-the-work-around-the-decision",[77],{"type":54,"value":78},"AI can do the work around the decision",{"type":49,"tag":50,"props":80,"children":81},{},[82],{"type":54,"value":83},"Most of what keeps an HR team busy is not deciding about people. It is\nbooking, replying, explaining, collecting, checking and reminding. That work\ncan be automated with peace of mind, and it is where most of the recoverable\nhours are.",{"type":49,"tag":50,"props":85,"children":86},{},[87],{"type":54,"value":88},"What calls for a different level of care is the decision itself: who moves on\nto the next stage, who gets promoted, who leaves, how someone's performance is\nassessed. AI can still help here, as long as the system is designed to support\na person who decides, and not to decide with a person signing underneath.",{"type":49,"tag":62,"props":90,"children":92},{"id":91},"where-ai-in-hr-is-straightforward",[93],{"type":54,"value":94},"Where AI in HR is straightforward",{"type":49,"tag":50,"props":96,"children":97},{},[98],{"type":54,"value":99},"There is a set of uses that, built properly, stay outside high risk and free\nup real time:",{"type":49,"tag":101,"props":102,"children":103},"ul",{},[104,116,126,136,146],{"type":49,"tag":105,"props":106,"children":107},"li",{},[108,114],{"type":49,"tag":109,"props":110,"children":111},"strong",{},[112],{"type":54,"value":113},"Interview scheduling.",{"type":54,"value":115}," Matching interviewer and candidate availability,\nproposing slots, rescheduling and sending reminders. It is logistics, not\nassessment.",{"type":49,"tag":105,"props":117,"children":118},{},[119,124],{"type":49,"tag":109,"props":120,"children":121},{},[122],{"type":54,"value":123},"Replying to candidates.",{"type":54,"value":125}," Confirming receipt, saying which stage the\nprocess has reached and answering questions about the role, the hours or the\nlocation, based on what the company has published.",{"type":49,"tag":105,"props":127,"children":128},{},[129,134],{"type":49,"tag":109,"props":130,"children":131},{},[132],{"type":54,"value":133},"Employee questions about internal policies.",{"type":54,"value":135}," How many days of holiday are\nleft, how to claim back an expense, what the handbook says about remote\nworking. An assistant that answers from company documents, cites its source\nand hands over to a person when the answer depends on a decision.",{"type":49,"tag":105,"props":137,"children":138},{},[139,144],{"type":49,"tag":109,"props":140,"children":141},{},[142],{"type":54,"value":143},"Onboarding.",{"type":54,"value":145}," Collecting documents before day one, preparing the contract\nfor signature, requesting system access, scheduling mandatory training and\ntracking what each new starter still has outstanding.",{"type":49,"tag":105,"props":147,"children":148},{},[149,154,156,163],{"type":49,"tag":109,"props":150,"children":151},{},[152],{"type":54,"value":153},"Document processing.",{"type":54,"value":155}," Reading certificates, declarations and occupational\nhealth reports, extracting expiry dates and flagging them before they lapse,\nwith the same validation logic we use for ",{"type":49,"tag":157,"props":158,"children":160},"a",{"href":159},"/en/blog/ai-document-processing-invoices-orders",[161],{"type":54,"value":162},"invoices and orders read by\nAI",{"type":54,"value":164},".",{"type":49,"tag":50,"props":166,"children":167},{},[168],{"type":54,"value":169},"These uses are not free of rules: an assistant has to make clear that it is\nautomated, and personal data remains subject to the GDPR. But these are\nobligations that get solved at the design stage without making the project\nany heavier.",{"type":49,"tag":62,"props":171,"children":173},{"id":172},"where-high-risk-begins",[174],{"type":54,"value":175},"Where high risk begins",{"type":49,"tag":50,"props":177,"children":178},{},[179],{"type":54,"value":180},"Annex III of the AI Act lists the areas in which an AI system is, by default,\nhigh risk. Employment is one of them, and the wording is broader than most\npeople expect: systems used to recruit, filter applications and assess\ncandidates, to decide on promotions or the termination of contracts, to\nallocate tasks based on personal traits and to monitor performance.",{"type":49,"tag":182,"props":183,"children":184},"table",{},[185,209],{"type":49,"tag":186,"props":187,"children":188},"thead",{},[189],{"type":49,"tag":190,"props":191,"children":192},"tr",{},[193,199,204],{"type":49,"tag":194,"props":195,"children":196},"th",{},[197],{"type":54,"value":198},"Use",{"type":49,"tag":194,"props":200,"children":201},{},[202],{"type":54,"value":203},"Likely classification",{"type":49,"tag":194,"props":205,"children":206},{},[207],{"type":54,"value":208},"What it implies",{"type":49,"tag":210,"props":211,"children":212},"tbody",{},[213,232,250,267,285,302],{"type":49,"tag":190,"props":214,"children":215},{},[216,222,227],{"type":49,"tag":217,"props":218,"children":219},"td",{},[220],{"type":54,"value":221},"Scheduling interviews and sending reminders",{"type":49,"tag":217,"props":223,"children":224},{},[225],{"type":54,"value":226},"Outside high risk",{"type":49,"tag":217,"props":228,"children":229},{},[230],{"type":54,"value":231},"Good practice and GDPR",{"type":49,"tag":190,"props":233,"children":234},{},[235,240,245],{"type":49,"tag":217,"props":236,"children":237},{},[238],{"type":54,"value":239},"Internal policy assistant",{"type":49,"tag":217,"props":241,"children":242},{},[243],{"type":54,"value":244},"Limited risk",{"type":49,"tag":217,"props":246,"children":247},{},[248],{"type":54,"value":249},"Disclose it is AI, cite sources",{"type":49,"tag":190,"props":251,"children":252},{},[253,258,262],{"type":49,"tag":217,"props":254,"children":255},{},[256],{"type":54,"value":257},"Extracting data from onboarding documents",{"type":49,"tag":217,"props":259,"children":260},{},[261],{"type":54,"value":226},{"type":49,"tag":217,"props":263,"children":264},{},[265],{"type":54,"value":266},"GDPR, access control",{"type":49,"tag":190,"props":268,"children":269},{},[270,275,280],{"type":49,"tag":217,"props":271,"children":272},{},[273],{"type":54,"value":274},"Ranking or scoring applications",{"type":49,"tag":217,"props":276,"children":277},{},[278],{"type":54,"value":279},"High risk",{"type":49,"tag":217,"props":281,"children":282},{},[283],{"type":54,"value":284},"Oversight, records, testing, disclosure",{"type":49,"tag":190,"props":286,"children":287},{},[288,293,297],{"type":49,"tag":217,"props":289,"children":290},{},[291],{"type":54,"value":292},"Supporting promotion or appraisal decisions",{"type":49,"tag":217,"props":294,"children":295},{},[296],{"type":54,"value":279},{"type":49,"tag":217,"props":298,"children":299},{},[300],{"type":54,"value":301},"The same, plus informing employees",{"type":49,"tag":190,"props":303,"children":304},{},[305,310,315],{"type":49,"tag":217,"props":306,"children":307},{},[308],{"type":54,"value":309},"Inferring emotions in interviews or at work",{"type":49,"tag":217,"props":311,"children":312},{},[313],{"type":54,"value":314},"Prohibited",{"type":49,"tag":217,"props":316,"children":317},{},[318],{"type":54,"value":319},"Not done",{"type":49,"tag":50,"props":321,"children":322},{},[323],{"type":54,"value":324},"There is one detail worth knowing. The regulation allows a system listed in\nAnnex III not to be treated as high risk when it performs a narrow procedural\nor preparatory task without materially influencing the decision. That\nexception, however, does not apply when the system profiles people, and\nscoring candidates on their fit for a role is hard to describe as anything\nelse. The\nbutton from the start of this article sits on the wrong side of that line.",{"type":49,"tag":50,"props":326,"children":327},{},[328],{"type":54,"value":329},"Several practices have been prohibited since 2 February 2025, among them\ninferring people's emotions in the workplace, except for medical or safety\nreasons. Tools that claim to read enthusiasm or nerves in a video interview\nare out of the question.",{"type":49,"tag":50,"props":331,"children":332},{},[333],{"type":54,"value":334},"As for the timeline, the high-risk obligations were due on 2 August 2026. The\nEuropean simplification package known as the Digital Omnibus, published as\nRegulation (EU) 2026/1744 and in force since 27 July 2026, pushed them back,\nfor Annex III systems, to 2 December 2027. Since the timeline has already moved\nonce, confirm the date with your legal team before fixing any plans. The\npostponement does not suspend anything else: the duty to tell people they are\ninteracting with an AI system has applied since 2 August 2026, and the GDPR\nalready gives people the right not to be subject to solely automated decisions\nthat significantly affect them, citing e-recruiting with no human involvement\nas an example. In Portugal, the Labour\nCode has required employers since 2023 to disclose the parameters and criteria\nof algorithms that affect access to and retention of employment.",{"type":49,"tag":50,"props":336,"children":337},{},[338],{"type":54,"value":339},"One point that often comes as a surprise: a company that commissions a\ncandidate assessment system and puts it into service for its own use can end\nup being treated as the provider of that system, with documentation and risk\nmanagement obligations far heavier than those of a company that merely uses\none. That is one more reason to think twice before building an engine that\nscores people.",{"type":49,"tag":62,"props":341,"children":343},{"id":342},"where-ai-has-already-crept-in-without-anyone-deciding",[344],{"type":54,"value":345},"Where AI has already crept in without anyone deciding",{"type":49,"tag":50,"props":347,"children":348},{},[349],{"type":54,"value":350},"Many companies are already using AI in decisions about people without ever\nhaving decided to:",{"type":49,"tag":101,"props":352,"children":353},{},[354,359,364,369,374,379],{"type":49,"tag":105,"props":355,"children":356},{},[357],{"type":54,"value":358},"the recruitment software shows a score or an automatic ranking of\ncandidates;",{"type":49,"tag":105,"props":360,"children":361},{},[362],{"type":54,"value":363},"someone on the team pastes CVs into a general-purpose assistant and asks for\n\"the top five\";",{"type":49,"tag":105,"props":365,"children":366},{},[367],{"type":54,"value":368},"automatic filters remove applications before a person ever sees them;",{"type":49,"tag":105,"props":370,"children":371},{},[372],{"type":54,"value":373},"performance reports include indicators calculated from activity in company\nsystems, without employees knowing how they are worked out;",{"type":49,"tag":105,"props":375,"children":376},{},[377],{"type":54,"value":378},"nobody can say who approved switching on an AI feature in the HR software;",{"type":49,"tag":105,"props":380,"children":381},{},[382],{"type":54,"value":383},"rejected candidates get an automated reply and there is no way to find out\nwhy they were excluded.",{"type":49,"tag":62,"props":385,"children":387},{"id":386},"three-common-reactions-and-why-they-fail",[388],{"type":54,"value":389},"Three common reactions, and why they fail",{"type":49,"tag":50,"props":391,"children":392},{},[393,398],{"type":49,"tag":109,"props":394,"children":395},{},[396],{"type":54,"value":397},"Switch the module on and trust the vendor.",{"type":54,"value":399}," The vendor has its\nobligations, but the company using the system has its own: oversight by\ntrained people, records kept, candidates informed and, before a system like\nthis is used in the workplace, employees and their representatives informed\ntoo. None of that comes with the click.",{"type":49,"tag":50,"props":401,"children":402},{},[403,408],{"type":49,"tag":109,"props":404,"children":405},{},[406],{"type":54,"value":407},"Ban AI from HR altogether.",{"type":54,"value":409}," Understandable, and it leaves the team exactly\nwhere it was, answering hundreds of emails by hand about the status of an\napplication. Worse, the informal use of general-purpose tools carries on, only\nnow without any rules.",{"type":49,"tag":50,"props":411,"children":412},{},[413,418],{"type":49,"tag":109,"props":414,"children":415},{},[416],{"type":54,"value":417},"Have a person press the final button.",{"type":54,"value":419}," This is the answer that looks like\na solution and is not one. If the decision maker receives a ranked list, cannot\nsee why candidate 340 ended up in 340th place and has thirty minutes to finish\nscreening, the human involvement is decorative. A review that always confirms\nwhat the system proposed is not oversight, it is a signature.",{"type":49,"tag":62,"props":421,"children":423},{"id":422},"how-to-design-ai-that-supports-rather-than-decides",[424],{"type":54,"value":425},"How to design AI that supports rather than decides",{"type":49,"tag":50,"props":427,"children":428},{},[429],{"type":54,"value":430},"The difference between supporting and deciding does not lie in the language\nmodel, it lies in the architecture around it. Five decisions do almost all of\nthe work.",{"type":49,"tag":50,"props":432,"children":433},{},[434,439],{"type":49,"tag":109,"props":435,"children":436},{},[437],{"type":54,"value":438},"Criteria are set by people and kept under version control.",{"type":54,"value":440}," Before the\nsystem reads a single application, HR and the hiring manager define what\ncounts for that role: an HGV licence, experience with barcode scanners,\navailability for night shifts. Each criterion is written down, dated and tied\nto the role. If it changes midway through the process, the record shows when\nand by whom.",{"type":49,"tag":50,"props":442,"children":443},{},[444,449],{"type":49,"tag":109,"props":445,"children":446},{},[447],{"type":54,"value":448},"The system extracts evidence, it does not give marks.",{"type":54,"value":450}," Instead of an\noverall score, each application is presented criterion by criterion, with the\npassage from the CV that backs up each statement. \"Mentions three years as a\nforklift operator, 2021 to 2024, in the experience section.\" The decision maker\nsees what the system found and where, and can check it in seconds. When the\ninformation is not there, the system says it did not find it, rather than\ninferring it.",{"type":49,"tag":50,"props":452,"children":453},{},[454,459],{"type":49,"tag":109,"props":455,"children":456},{},[457],{"type":54,"value":458},"There is no automatic rejection.",{"type":54,"value":460}," The system can flag that an application\ndoes not mention a mandatory requirement, but the exclusion is made by a\nperson. Nothing leaves the process without a human decision attached to a\nname.",{"type":49,"tag":50,"props":462,"children":463},{},[464,469],{"type":49,"tag":109,"props":465,"children":466},{},[467],{"type":54,"value":468},"The decision is recorded with its reasoning.",{"type":54,"value":470}," Every move between stages\nstores who decided, when, with what information in front of them and why.\nThat is what makes it possible to answer a candidate who asks for an\nexplanation, an audit or a complaint, and it is also how you measure whether\nthe human review is real: if nobody ever disagrees with the system, there is\na problem worth investigating.",{"type":49,"tag":50,"props":472,"children":473},{},[474,479],{"type":49,"tag":109,"props":475,"children":476},{},[477],{"type":54,"value":478},"Bias testing before and after go-live.",{"type":54,"value":480}," Removing the name and the photo is\nnot enough, because some variables act as proxies: the postcode, the year a\ndegree was completed, which gives away age, or career breaks that coincide\nwith parental leave. We test with sets of applications built for the purpose\nand repeat the test whenever the criteria or the model change.",{"type":49,"tag":50,"props":482,"children":483},{},[484,486,492],{"type":54,"value":485},"On top of this come the information given to candidates, who know there is AI\nin the process and how to ask for a review by a person, and the data\nprotection impact assessment, carried out at the design stage rather than\nafterwards. The general principles we apply to any agent in production are in\nthe article on ",{"type":49,"tag":157,"props":487,"children":489},{"href":488},"/en/blog/gdpr-ai-act-before-ai-agents-production",[490],{"type":54,"value":491},"GDPR and the AI Act before putting AI agents into\nproduction",{"type":54,"value":493},"; the difference\nhere is that the bar goes up because the decision is about people.",{"type":49,"tag":50,"props":495,"children":496},{},[497],{"type":54,"value":498},"We do not give legal advice, and the final classification of each use is for\nthe company and its advisers to make. Our job is to build the system so that\nthis analysis finds what it needs: written criteria, complete records and\ndecisions that can be explained to a candidate, an employee or an inspector.",{"type":49,"tag":62,"props":500,"children":502},{"id":501},"how-we-work",[503],{"type":54,"value":504},"How we work",{"type":49,"tag":506,"props":507,"children":509},"h3",{"id":508},"_1-free-diagnosis",[510],{"type":54,"value":511},"1. Free diagnosis",{"type":49,"tag":50,"props":513,"children":514},{},[515],{"type":54,"value":516},"A conversation with no commitment to map where AI is already used in HR,\nofficially or not, and where the team loses the most hours. By the end, each\nuse is classified: straightforward, needs care, or best avoided.",{"type":49,"tag":506,"props":518,"children":520},{"id":519},"_2-design-and-criteria",[521],{"type":54,"value":522},"2. Design and criteria",{"type":49,"tag":50,"props":524,"children":525},{},[526],{"type":54,"value":527},"Together with the HR leadership we define the scope, the criteria for each\nprocess, who decides what and what gets recorded. This is also where the data\nprotection officer and, where there is one, the works council get involved.",{"type":49,"tag":506,"props":529,"children":531},{"id":530},"_3-first-what-frees-up-hours-without-risk",[532],{"type":54,"value":533},"3. First, what frees up hours without risk",{"type":49,"tag":50,"props":535,"children":536},{},[537],{"type":54,"value":538},"Employee portal, internal policy assistant, digital onboarding and\nscheduling. This is the phase with the fastest return, and the one that\nbuilds the data and access foundation everything else rests on.",{"type":49,"tag":506,"props":540,"children":542},{"id":541},"_4-decision-support-if-it-makes-sense",[543],{"type":54,"value":544},"4. Decision support, if it makes sense",{"type":49,"tag":50,"props":546,"children":547},{},[548],{"type":54,"value":549},"Only afterwards, and only if the volume justifies it, evidence extraction from\napplications, with human review, record keeping and bias testing designed in\nfrom the start.",{"type":49,"tag":506,"props":551,"children":553},{"id":552},"_5-production-and-ongoing-support",[554],{"type":54,"value":555},"5. Production and ongoing support",{"type":49,"tag":50,"props":557,"children":558},{},[559],{"type":54,"value":560},"Training for the people who use the system, which also serves the duty to\npromote AI literacy that the AI Act places on those who use it, close support\nthrough the first weeks and periodic review of the tests. The code and the\ninfrastructure belong to the company.",{"type":49,"tag":62,"props":562,"children":564},{"id":563},"reference-figures-and-timelines",[565],{"type":54,"value":566},"Reference figures and timelines",{"type":49,"tag":50,"props":568,"children":569},{},[570,572,578,580,585,587,592,594,600,602,607,609,614,616,622,624,629],{"type":54,"value":571},"The figures depend on the starting point. Within our ",{"type":49,"tag":157,"props":573,"children":575},{"href":574},"/en/solucoes/onboarding-digital",[576],{"type":54,"value":577},"digital\nonboarding",{"type":54,"value":579}," solution, an employee portal with\nholidays, expenses, documents and a personal area sits ",{"type":49,"tag":109,"props":581,"children":582},{},[583],{"type":54,"value":584},"between €15,000 and\n€25,000",{"type":54,"value":586},", rising to ",{"type":49,"tag":109,"props":588,"children":589},{},[590],{"type":54,"value":591},"between €25,000 and €40,000",{"type":54,"value":593}," when it includes digital\nonboarding of new starters with electronic signature. For ",{"type":49,"tag":157,"props":595,"children":597},{"href":596},"/en/solucoes/gestao-recursos-humanos-customizado",[598],{"type":54,"value":599},"custom HR\nmanagement",{"type":54,"value":601},", a portal with\ntime and attendance comes in ",{"type":49,"tag":109,"props":603,"children":604},{},[605],{"type":54,"value":606},"between €18,000 and €30,000",{"type":54,"value":608},", rotas and\nshifts with legal rules ",{"type":49,"tag":109,"props":610,"children":611},{},[612],{"type":54,"value":613},"between €30,000 and €50,000",{"type":54,"value":615}," (a concrete example is\n",{"type":49,"tag":157,"props":617,"children":619},{"href":618},"/en/blog/staff-management-hospitality",[620],{"type":54,"value":621},"shift management in hospitality",{"type":54,"value":623},"), and a complete\npeople platform, with appraisals and skills, ",{"type":49,"tag":109,"props":625,"children":626},{},[627],{"type":54,"value":628},"from €50,000",{"type":54,"value":164},{"type":49,"tag":50,"props":631,"children":632},{},[633,635,641],{"type":54,"value":634},"What moves the figure most is the number of systems to integrate (payroll,\ntime tracking, existing recruitment software) and the quality of the internal\ndocuments the assistant will rely on. A screening support component, where it\nis justified, adds its own record keeping and testing work, which should not\nbe cut to bring the price down. As an engineering plan, building a portal\nwith onboarding typically takes 8 to 12 weeks, and 12 to 18 weeks when it\nincludes attendance, rotas or appraisals. What makes these timelines vary is\nexplained in ",{"type":49,"tag":157,"props":636,"children":638},{"href":637},"/en/blog/how-long-a-custom-software-project-takes",[639],{"type":54,"value":640},"how long a custom software or AI project\ntakes",{"type":54,"value":164},{"type":49,"tag":62,"props":643,"children":645},{"id":644},"when-the-answer-is-not-to-build",[646],{"type":54,"value":647},"When the answer is not to build",{"type":49,"tag":50,"props":649,"children":650},{},[651,653,659],{"type":54,"value":652},"If the company receives a few dozen applications a month, AI screening does\nnot pay for itself and adds obligations with no return. If all you are after\nis interview scheduling and automatic replies to candidates, off-the-shelf\nrecruitment software is usually enough, and that is what we recommend. The\nbroader comparison is in the article on ",{"type":49,"tag":157,"props":654,"children":656},{"href":655},"/en/blog/custom-ai-agents-vs-generic-tools",[657],{"type":54,"value":658},"custom AI agents versus generic\ntools",{"type":54,"value":164},{"type":49,"tag":50,"props":661,"children":662},{},[663],{"type":54,"value":664},"And if the goal is an automatic score that decides who gets interviewed, our\nhonest answer is that we do not build it that way, because it shifts onto the\ncompany a risk that the time saved does not make up for.",{"type":49,"tag":62,"props":666,"children":668},{"id":667},"five-questions-for-the-hr-director",[669],{"type":54,"value":670},"Five questions for the HR director",{"type":49,"tag":101,"props":672,"children":673},{},[674,679,684,689,694],{"type":49,"tag":105,"props":675,"children":676},{},[677],{"type":54,"value":678},"How many applications do you receive a month, and how many hours does the\nteam spend just replying and scheduling?",{"type":49,"tag":105,"props":680,"children":681},{},[682],{"type":54,"value":683},"Which AI features are currently switched on in your HR software, and who\napproved them?",{"type":49,"tag":105,"props":685,"children":686},{},[687],{"type":54,"value":688},"Are your internal policies written down and up to date, or do they live in\nthe heads of two people?",{"type":49,"tag":105,"props":690,"children":691},{},[692],{"type":54,"value":693},"Who in the company is responsible for data protection and for employee\nrelations?",{"type":49,"tag":105,"props":695,"children":696},{},[697],{"type":54,"value":698},"If a candidate asked you tomorrow why they were rejected, could you answer?",{"type":49,"tag":50,"props":700,"children":701},{},[702,704,710],{"type":54,"value":703},"If the last question had no answer, that is the place to start. If you want to\nunderstand where AI can help your HR team without putting the company in a\ncategory it never chose, you can ",{"type":49,"tag":157,"props":705,"children":707},{"href":706},"/en/contacto",[708],{"type":54,"value":709},"request a free diagnosis",{"type":54,"value":711},". In the first conversation, with no commitment, we go through the uses already in place and the ones worth adding, and classify each of them with you.",{"title":713,"searchDepth":714,"depth":714,"links":715},"",3,[716,718,719,720,721,722,723,724,731,732,733],{"id":64,"depth":717,"text":67},2,{"id":75,"depth":717,"text":78},{"id":91,"depth":717,"text":94},{"id":172,"depth":717,"text":175},{"id":342,"depth":717,"text":345},{"id":386,"depth":717,"text":389},{"id":422,"depth":717,"text":425},{"id":501,"depth":717,"text":504,"children":725},[726,727,728,729,730],{"id":508,"depth":714,"text":511},{"id":519,"depth":714,"text":522},{"id":530,"depth":714,"text":533},{"id":541,"depth":714,"text":544},{"id":552,"depth":714,"text":555},{"id":563,"depth":717,"text":566},{"id":644,"depth":717,"text":647},{"id":667,"depth":717,"text":670},"markdown","content:en:blog:ai-recruitment-hr-ai-act-high-risk.md","content","en/blog/ai-recruitment-hr-ai-act-high-risk.md","en/blog/ai-recruitment-hr-ai-act-high-risk","md",[741,748,757,765,772,780,788,795,804,811,818,825,833,840,847,855,861,867,874,881,888,895,901,909,915,922,929,936,944,952,959,966,972,979,986,993,999,1007,1014,1021,1028,1035,1042,1050],{"_path":742,"title":743,"description":744,"date":13,"image":745,"tags":746},"/blog/custom-software-quote-what-it-should-include","What a Custom Software Quote Should Include: A Line-by-Line Example","The line items a custom software quote should break down, a line-by-line example, the costs left out and the warning signs to check before signing.","/images/blog/orcamento-software-medida.jpg",[747,22],"Costs & Decision",{"_path":749,"title":750,"description":751,"date":752,"image":753,"tags":754},"/blog/real-estate-crm-software-portugal-how-to-choose","Real Estate CRM Software in Portugal: How to Choose the Right One for Your Agency","Portals, listings, viewings, commissions and IMPIC obligations: what a real estate CRM has to handle in Portugal and when a custom-built one pays off.","2026-11-09","/images/blog/crm-imobiliario-portugal.jpg",[755,756,22],"Real Estate","CRM",{"_path":758,"title":759,"description":760,"date":761,"image":762,"tags":763},"/blog/is-your-company-data-ready-for-ai","Is Your Company Data Ready for AI? How to Tell Before You Invest","What each type of AI project asks of your data, an eight-question self-assessment, and why you do not need a data lake before your first AI project.","2026-10-29","/images/blog/dados-prontos-ia.jpg",[19,764,747,22],"Business Intelligence",{"_path":766,"title":767,"description":768,"date":769,"image":770,"tags":771},"/blog/ai-cash-flow-treasury-forecasting","Cash Flow Forecasting: Stop Running Treasury on Bank Statements and a Friday Spreadsheet","Consolidated multi-bank position, automatic reconciliation and a 13-week cash flow forecast. Where AI genuinely helps and where simple statistics are enough.","2026-10-24","/images/blog/previsao-tesouraria.jpg",[764,19,21,747],{"_path":773,"title":774,"description":775,"date":776,"image":777,"tags":778},"/blog/ai-clinical-documentation-consultations","AI for Clinical Documentation: Fewer Hours Writing, Without Losing Control of the Note","How AI transcribes and summarises consultations, drafts letters and writes into clinical software, with the clinician signing off. GDPR, MDR and costs.","2026-10-19","/images/blog/ia-documentacao-clinica.jpg",[779,19,20],"Healthcare",{"_path":781,"title":782,"description":783,"date":784,"image":785,"tags":786},"/blog/fleet-rent-a-car-management-software-ai","Custom Rent-a-Car Software: Real Fleet Status, AI Inspections and Dynamic Pricing","Real availability, inspection photos compared by AI, fines passed to the renter on time and occupancy-based pricing: what a rent-a-car operation needs.","2026-10-14","/images/blog/frotas-rent-a-car.jpg",[787,22,19],"Logistics",{"_path":789,"title":790,"description":791,"date":792,"image":793,"tags":794},"/blog/field-service-management-software-technical-teams","Software for Field Technical Teams: From the Call to the Invoice Without Paper","Work orders, an offline app, parts in the van, contracts with SLAs and an invoice at close-out: what a system for field technicians really has to solve.","2026-10-03","/images/blog/equipas-tecnicas-terreno.jpg",[22,21,19],{"_path":796,"title":797,"description":798,"date":799,"image":800,"tags":801},"/blog/custom-erp-wholesale-distribution","Custom ERP for Wholesalers and Distributors: Where the Margin Goes","Customer pricing, volume rebates, WhatsApp orders and credit limits: where an off-the-shelf ERP fails a wholesaler, where AI helps and what custom costs.","2026-09-26","/images/blog/erp-distribuicao-grossista.jpg",[787,802,22,803],"ERP","B2B Sales",{"_path":805,"title":806,"description":807,"date":808,"image":809,"tags":810},"/blog/ai-document-processing-invoices-orders","Invoices, Purchase Orders and Delivery Notes in PDF: How AI Reads Them and Posts Them to the ERP Without Silent Errors","How AI reads supplier invoices, purchase orders and delivery notes from email, checks them against the ERP and leaves only exceptions to your team.","2026-09-19","/images/blog/ia-processamento-documentos.jpg",[21,19,802,747],{"_path":812,"title":813,"description":814,"date":815,"image":816,"tags":817},"/blog/modernise-legacy-software-without-stopping-operations","The Old System Nobody Wants to Touch: How to Modernise It Without Stopping Operations","Access, Visual Basic, Delphi or a 15-year-old ERP: how to replace old software one module at a time, migrate data in parallel and keep the business running.","2026-09-12","/images/blog/modernizar-software-legado.jpg",[22,802,747],{"_path":819,"title":820,"description":821,"date":822,"image":823,"tags":824},"/blog/ai-prototype-to-production","Your Team Built an AI Prototype. What It Takes to Get to Production","More and more teams build applications with ChatGPT, Codex or Claude Code. What changes when that software has to be secure, reliable and able to scale to production.","2026-09-05","/images/blog/ai-prototype-to-production.jpg",[22,19,20,747],{"_path":826,"title":827,"description":828,"date":829,"image":830,"tags":831},"/blog/integrate-ai-business-processes-digital-brain","How to Integrate AI into Business Processes Without Adding Another Isolated Tool","Learn how a Digital Brain connects documents, ERP, CRM and email to turn AI into real work, with security, control and measurable return.","2026-08-26","/images/blog/cerebro-digital.jpg",[19,832,21,747],"AI Agents",{"_path":834,"title":835,"description":836,"date":837,"image":838,"tags":839},"/blog/how-long-a-custom-software-project-takes","How Long a Custom Software or AI Project Actually Takes","Real timelines by project type, what happens inside each phase, and the four things on the client side that delay projects more than any technical decision.","2026-08-22","/images/blog/prazos-projeto-software-medida.jpg",[747,22,19,21],{"_path":841,"title":842,"description":843,"date":844,"image":845,"tags":846},"/blog/add-ai-to-existing-business-software","Adding AI to the Software Your Company Already Uses, Without Replacing It","How AI connects to the ERP, the CRM and the files you already have, what changes when a system has no ready-made connection, and what each scenario actually costs.","2026-08-16","/images/blog/integrar-ia-software-existente.jpg",[19,22,802,747],{"_path":848,"title":849,"description":850,"date":851,"image":852,"tags":853},"/blog/how-to-choose-an-ai-agent-partner","How to Choose Who Builds Your Company AI Agents","Seven criteria to separate those who deliver from those who give a good demo: code ownership, fixed scope, what happens when the AI is wrong, and where the data lives.","2026-08-09","/images/blog/agentes-ia-empresariais-suporte-b2b.jpg",[832,747,854],"B2B Portals",{"_path":856,"title":857,"description":858,"date":859,"image":852,"tags":860},"/blog/gdpr-ai-act-before-ai-agents-production","GDPR and the AI Act: What to Check Before Putting an AI Agent in Production","What the GDPR and the AI Act actually require from a company using AI agents, the five points that cover most cases, and the mistake that costs the most.","2026-08-02",[20,832],{"_path":862,"title":863,"description":864,"date":865,"image":830,"tags":866},"/blog/how-much-ai-costs-company-portugal","How Much Does It Cost to Implement AI in a Company in Portugal in 2026","Real price ranges for AI agents, internal platforms and custom software in Portugal, what drives the figure up, and the running cost that never appears in proposals.","2026-07-26",[747,19,832],{"_path":868,"title":869,"description":870,"date":871,"image":872,"tags":873},"/blog/custom-erp-crm-vs-primavera-phc-sap","Custom ERP and CRM or Primavera, PHC, ARTSOFT and SAP: How to Decide","Primavera, PHC, ARTSOFT and SAP fit most companies. The three signs the product stopped fitting, and why replacing everything rarely pays off.","2026-07-19","/images/blog/agribusiness-erp.jpg",[802,756,747],{"_path":875,"title":876,"description":877,"date":878,"image":879,"tags":880},"/blog/custom-ai-agents-vs-generic-tools","Custom AI Agents or Generic Tools: When Building Is Worth It","ChatGPT and Copilot solve individual productivity. When the process needs company data and has to act in your systems, the maths changes. Criteria for deciding.","2026-07-12","/images/blog/automacao-vendas-ia.jpg",[832,19,747],{"_path":882,"title":883,"description":884,"date":885,"image":886,"tags":887},"/blog/custom-real-estate-crm-high-ticket-deals","Custom Real Estate CRM: The Key to Closing High-Ticket Deals","Discover why generic CRMs fail in the luxury market and how a custom-built real estate CRM with AI integration accelerates closing high-ticket deals.","2026-06-08","/images/blog/crm-real-estate.jpg",[756,755,19],{"_path":889,"title":890,"description":891,"date":892,"image":893,"tags":894},"/blog/legal-erp-integrated-crm-lead-capture","From Followers to Clients: How a Legal ERP with Integrated CRM Transforms Lead Capture","What changes when client acquisition and case management live in the same system, and why most of the enquiries generated on social channels are lost before they ever reach a fee proposal.","2026-05-21","/images/blog/erp-juridico-crm.jpg",[22,756,21],{"_path":896,"title":897,"description":898,"date":899,"image":879,"tags":900},"/blog/ai-sales-automation","How AI Automation Shortens Sales Cycles and Boosts Conversion Rates","Discover how AI sales automation eliminates manual administrative work, accelerates high-value B2B deals, and transforms your CRM.","2026-05-14",[19,21,803],{"_path":902,"title":903,"description":904,"date":905,"image":906,"tags":907},"/blog/e-commerce-checkout-optimization","Why Your Online Store is Losing Sales at Checkout (and How to Fix It)","Discover why an outdated e-commerce platform turns customers away at checkout and how modern payment integrations immediately increase your conversion rate.","2026-05-12","/images/blog/otimizacao-checkout-e-commerce.jpg",[908,22],"Retail & E-commerce",{"_path":910,"title":911,"description":912,"date":913,"image":830,"tags":914},"/blog/digital-brain-institutional-memory-smes","What if Your Company Had a Digital Brain? The Revolution of Institutional Memory in SMEs","Discover how a Digital Brain powered by Private AI can centralize company knowledge, eliminate the Bus Factor, and scale your operations securely.","2026-05-08",[19,21],{"_path":916,"title":917,"description":918,"date":919,"image":920,"tags":921},"/blog/ai-chatbot-qualifies-leads-24h","Multiply Your Sales: How an AI Chatbot Qualifies Leads 24 Hours a Day","Discover how a custom Artificial Intelligence chatbot can interact, qualify high-value potential clients, and schedule meetings while your team sleeps.","2026-05-02","/images/blog/chatbot-ia-qualifica-leads.jpg",[19,803,21],{"_path":923,"title":924,"description":925,"date":926,"image":927,"tags":928},"/blog/custom-wms-ai-inventory-management","Zero-Touch Inventory Management: How Custom WMS & AI Reduce Stockouts","Discover how a custom Warehouse Management System (WMS) powered by AI reduces stockouts and automates replenishments in retail and distribution operations.","2026-04-24","/images/blog/wms-medida-ia-gestao-inventario.jpg",[908,19,787],{"_path":930,"title":931,"description":932,"date":933,"image":934,"tags":935},"/blog/custom-erp-consulting-firms-project-management","Ending Consultancy Chaos: 10 SaaS Tools vs 1 Custom ERP","Discover why dozens of SaaS tools destroy consultancy profit margins and how a custom ERP safely restores holistic financial control.","2026-04-20","/images/blog/finance-automation.jpg",[22,802],{"_path":937,"title":938,"description":939,"date":940,"image":941,"tags":942},"/blog/custom-eqms-manufacturing-quality-control","Paperless Quality Control: The Value of Custom e-QMS in Manufacturing","Discover how a custom e-QMS replaces manual factory floor processes and accelerates audits in manufacturing.","2026-04-17","/images/blog/eqms-sistema-controlo-qualidade-industria.jpg",[943,22,19],"Manufacturing",{"_path":945,"title":946,"description":947,"date":948,"image":949,"tags":950},"/blog/law-firm-management-software-ai","Law Firm Management Software: How AI Solves the Problem of Slow Contract Analysis","Discover how a custom ERP and management software for law firms integrates AI to analyze contracts and scale billable hours.","2026-04-15","/images/blog/software-gestao-advogados-ia.jpg",[19,22,951],"Legal Sector",{"_path":953,"title":954,"description":955,"date":956,"image":957,"tags":958},"/blog/custom-crm-vs-salesforce","What Is a Custom CRM and Why Salesforce Might Be Holding Your Business Back","The signs a generic CRM is no longer enough, what changes with a custom system, and how licence costs compare against ownership over a three to five year horizon.","2026-04-10","/images/blog/crm-medida-vs-salesforce.jpg",[756,803,22,747],{"_path":960,"title":961,"description":962,"date":963,"image":964,"tags":965},"/blog/custom-vendor-portals-procurement","Custom Vendor Portals: The Secret Strategy for a Resilient Supply Chain","Discover how bespoke Vendor Portals guarantee efficiency, rock-solid compliance, and total management control across complex procurement operations.","2026-04-09","/images/blog/portais-fornecedores-medida-procurement.jpg",[854,787],{"_path":967,"title":968,"description":969,"date":970,"image":852,"tags":971},"/blog/enterprise-ai-agents-b2b-support","Enterprise AI Agents: Reimagining Complex B2B Customer Support","How AI Agents integrated into your ERP transform B2B support, resolving complex queries without human intervention and scaling operations.","2026-04-08",[19,21],{"_path":973,"title":974,"description":975,"date":976,"image":977,"tags":978},"/blog/custom-erp-enterprise-groups-holdings","Multi-Company Management Without Chaos: The Strategic Advantage of a Custom ERP","Discover how Enterprise Groups and Holdings are abandoning rigid ERPs in favor of custom software to centralize management and scale profitably.","2026-04-06","/images/blog/erp-medida-grupos-empresariais-holdings.jpg",[22,802],{"_path":980,"title":981,"description":982,"date":983,"image":984,"tags":985},"/blog/traditional-mes-ai-smart-system","How to Transform Your Traditional MES into an AI-Driven Smart System","Passive MES systems bottleneck factory efficiency. Learn how to integrate Artificial Intelligence for predictive maintenance and near zero downtime.","2026-04-01","/images/blog/tradicional-mes-ai-smart-system.jpg",[943,19,22],{"_path":987,"title":988,"description":989,"date":990,"image":991,"tags":992},"/blog/custom-erp-vs-saas-logistics","Custom ERP vs SaaS: What Works Best for Logistics Operations?","Generic SaaS logistics platforms create bottlenecks. Learn why a custom ERP transforms distribution scalability and lowers long-term operational costs.","2026-03-23","/images/blog/erp-logistics.jpg",[787,802,22,747],{"_path":994,"title":995,"description":996,"date":997,"image":934,"tags":998},"/blog/backoffice-financial-automation","Backoffice Automation in Financial Services: Security and Efficiency","Manual approvals and document handling slow responses and create compliance risks. Discover how financial backoffice automation transforms operations.","2026-03-16",[21,20],{"_path":1000,"title":1001,"description":1002,"date":1003,"image":1004,"tags":1005},"/blog/end-of-excel-construction","The End of Excel in Construction: Centralizing Project Management","Budgets and planning scattered across spreadsheets cause massive financial overruns. Learn how a custom ERP safely centralizes construction management.","2026-03-09","/images/blog/excel-construction.jpg",[1006,802,22],"Construction",{"_path":1008,"title":1009,"description":1010,"date":1011,"image":1012,"tags":1013},"/blog/predictive-maintenance-manufacturing","Predictive Maintenance in Manufacturing: Using AI to Prevent Downtime","Unexpected assembly line failures destroy margins. Discover how predictive maintenance software uses AI to prevent breakdowns before they occur.","2026-03-02","/images/blog/predictive-maintenance.jpg",[943,19],{"_path":1015,"title":1016,"description":1017,"date":1018,"image":1019,"tags":1020},"/blog/ai-crm-real-estate","How AI is Transforming Lead Qualification in Real Estate CRMs","Real estate teams waste hours on unqualified leads. Discover how a custom CRM with AI agents eliminates manual prospecting and boosts conversion.","2026-02-23","/images/blog/ai-real-estate.jpg",[755,756,19,21],{"_path":1022,"title":1023,"description":1024,"date":1025,"image":1026,"tags":1027},"/blog/b2b-custom-extranet-portals","B2B Extranets: Retaining Enterprise Clients with Custom Portals","PDF and email ordering generate B2B friction. Discover how bespoke Extranet Portals streamline corporate sales and secure long-term client relationships.","2026-02-16","/images/blog/b2b-extranet.jpg",[854,943,21],{"_path":1029,"title":1030,"description":1031,"date":1032,"image":1033,"tags":1034},"/blog/custom-bi-dashboards-retail","Custom Business Intelligence Dashboards for Multi-Store Retail","Data silos between e-commerce, retail, and warehouses destroy margins. Learn how custom BI Dashboards centralize operations and drive decisions.","2026-02-09","/images/blog/retail-dashboard.jpg",[908,764,22],{"_path":1036,"title":1037,"description":1038,"date":1039,"image":872,"tags":1040},"/blog/agribusiness-custom-erp","Scaling Agribusiness: Why You Should Avoid Generic Management Software","Agricultural seasonality and traceability collide with traditional ERPs. Discover how custom software fully digitalizes agribusiness operations.","2026-02-02",[1041,802,22,943],"Agribusiness",{"_path":1043,"title":1044,"description":1045,"date":1046,"image":1047,"tags":1048},"/blog/staff-management-hospitality","Staff Management in Hospitality: Ending Roster Disorganization","Rosters in Excel and WhatsApp groups cost you shift coverage, maintenance traceability and GDPR compliance. What changes with a custom internal portal.","2026-01-26","/images/blog/hospitality-staff.jpg",[1049,854,21],"Hospitality",{"_path":1051,"title":1052,"description":1053,"date":1054,"image":1055,"tags":1056},"/blog/ai-clinical-scheduling","Clinical Scheduling Automation: Reducing No-Shows with AI Agents","Uncommunicated patient no-shows destroy daily clinic profitability. Learn how custom clinic scheduling software powered by AI actively eliminates friction.","2026-01-19","/images/blog/ai-clinic.jpg",[779,19,21],1794655862654]