Highlights from IAMPHENOM Europe 2026
Over two inspiring days in Paris, HR and business leaders came together at IAMPHENOM Europe to learn how global organisations are driving business outcomes with AI, agents, and automation across hiring, development, and retention.
Throughout 30+ main stage and breakout sessions, one overarching message stood out: the future of HR isn't just about adopting AI. It's about tailoring AI to solve an organisation’s most critical talent challenges and move the workforce forward.
Read on for the biggest takeaways from the event, including the latest agentic enhancements and service innovations, results from customers across Europe, and practical playbooks for scaling AI in your own organisation.
1. AI in HR needs both discipline and freedom
During the welcome HR Industry Keynote, our CEO and co-founder Mahe Bayireddi named the tension every HR leader feels today: we can be excited and anxious about AI at the same time. CEOs and CIOs want HR to hire faster, develop better and retain longer with less.
But enterprises are caught between two worlds: chaos that cannot be controlled, and control that cannot be harnessed. Rigid standardisation keeps the business safe, but slow. Unlimited experimentation moves fast but breaks things. Neither is the answer on its own.
"What we need is an architecture that gives us the freedom to adapt and the discipline to deliver," said Bayireddi.
Take French cooking, for example. The French gave the world a knowledge system that has inspired generations of cooking: codified kitchen stations, sauces, and repeatable standards that any professional kitchen still recognises today. That discipline was then scaled through cookbooks and television, freed up by a lighter, more personal style of cooking, and evolved as cultures blended French technique with their own. That journey took 200 years. HR has to cover the same ground in about five.
That is how we think about our Applied AI platform. Our core platform, built over a decade, delivers discipline and scale. The Hypercell and WorkOps add the freedom and evolution enterprises need to take on new, fast-growing work without putting the core at risk. The goal is to preserve the processes that run daily operations and compliance, while giving teams the flexibility to go after what comes next.
Remaining open to change is key. "It's not up to the company to determine whether it wants to use AI or not. It's the environment which dictates it," Bayireddi added.
2. Focus on AI outcomes, not activity
The purpose of AI for every enterprise is not to simply produce more work. It’s to achieve a desired outcome for the business. For HR, that can be a multitude of goals, such as reducing time to screen candidates, connecting employees with personalised career paths faster, and surfacing skills and task-level intelligence to inform workforce decisions more efficiently.
Customer examples showcased how AI, agents, and automation applied to a specific business problem:
A global science and technology company saved approximately €38 million through internal hires and €58 million through growth assignments since launching its talent marketplace.
At a global construction and services group, employees who use its internal mobility platform are five times more likely to make an internal move, and leavers are down 3%.
At a European home security company, field sales candidates screened by the Voice Screening Agent reached the interview stage at 6.9%, compared with 3.6% for roles without screening, and moved from application to interview two days faster.
A leading airline used a two-day AI Bootcamp to build a single English proficiency standard for safety-critical roles, then made 94 offers in one recruitment drive.
Phenom helps enterprises succeed by defining the outcome, applying the AI harness (or Hypercell context across the role, function, workflow, industry, and geography), and delivering the right experience. "If one thing is missing, the whole thing is destroyed," Bayireddi said. A well-defined outcome with a poor experience delivers nothing. A great experience without trust delivers nothing either.
Every Hypercell starts with a clearly defined outcome, orchestrates the work to reach it, is backed by an ontology and carries a measurable value. Each one gets its own harness with a mix of memory, context, model routing, agents, data and evaluations, because the right combination changes from one business to the next.
Evaluations then check that every part works as it should. At Phenom, we have embedded dozens of HR practitioners, former recruiters and HR leaders, alongside our AI and ML engineers to build evaluation data sets by industry and location. Like a head chef tasting every plate before it leaves the kitchen, evaluations hold quality to a global standard while adapting to local context.
With this level of precision, enterprises are equipped with the right AI, agents, and automation to support the specific use case and drive the right outcome.
One role, three markets, three workflows
Here’s what this level of precision looks like in practice: a quality control microbiologist might appear to be the same job, but the market context can change how one approaches hiring:
In a shortage market, where every pharma and biotech employer competes for the same people, the workflow searches internally and externally at once, keeps screening light, and schedules immediately.
In a balanced market near strong universities, it runs at a standard pace and prioritises quality.
In a market with a surplus of graduates, it shifts to behavioural assessments, hiring events and cohorts that start and train together.
Our Orchestration Engine recognises each market and adapts so no recruiter has to make a gut call or configure three workflows by hand. Think of it as the floor manager of a factory: automation handles the predictable, agents handle defined tasks, and orchestration handles the unexpected. If a hiring manager misses two interviews in a row, the Orchestration Engine reroutes.
It runs on policies written in plain language, such as "senior manager hires must be interviewed by a department head" or "this role goes internal for the first 30 days." Every decision, policy, and case escalated to a person are recorded.
With this level of precision, enterprises are equipped with the right AI, agents, and automation to drive the right outcome for every use case.
3. Connect HR work to what the business needs most
Traditional hiring systems have a major blind spot: they don't understand the business. The system will know a quality control role is open at a production site, but it won't know if that site is the most critical to the company's growth, or that one lab there is the bottleneck for every product it ships. This is the type of business context strategic leaders know instinctively.
The Phenom Ontology brings together data about your workforce and business, providing deeper insights to make more informed talent decisions. An ontology built for a science and technology company, for example, combines structured data (e.g., annual reports, quarterly results, details about 62,000 employees across 65 countries and 100+ production sites) with unstructured knowledge gathered into a living wiki. From that, the system identifies where the company is growing, and which roles and sites matter most.
The same company began to put this data into action by addressing internal mobility challenges. Employees told leadership they wanted to see a career, not just their current job. The team mapped more than 60,000 employees to 220 role clusters and roughly 860 critical skills. Its internal talent marketplace now has 85% adoption and more than 1,500 roles have been filled internally, with an estimated savings of roughly €38 million from internal hires and €58 million from short-term growth assignments since launch. The key was making mobility a business priority backed by the CEO, not an HR initiative.
Defining roles by the work, not the title
Ask three managers which skills a role needs and you get three answers. Ask them which tasks it involves and they agree in a minute. Task intelligence builds on that. An AI interviewer holds a 15-minute conversation with a lab manager and turns it into a structured task profile, with skills, proficiency levels, and the automation potential of each task. Run across every manager, it builds a task and skills architecture at scale.
That makes it possible to compare work across the whole organisation. Roles that share most of their tasks with a hard-to-fill job surface instantly. The Career Coach Agent then builds each interested employee a plan to close skills gaps, connects them with mentors, and books training in their calendar.
Mobility that reaches every employee
Customers further demonstrated how internal-first strategies can scale. A global flavour and fragrance company saw 67% of its pilot employees build a profile, and 65% said they would recommend the platform to a colleague. Teams whose managers actively engaged saw 74% adoption, signalling that managers carry mobility as much as technology does.
A European construction and services group with 154,000 people across more than 80 countries saw nearly 16,000 employees use its internal mobility platform in 12 months, generating 8,800 referrals. Employees who use it are five times more likely to make an internal move.
And with almost half of employees working on sites, AI agents also help build profiles, validate skills, and coach careers by voice in French and English.
Planning for talent you can't hire
Workforce planning is moving just as fast. When a manufacturer announces a new plant, the ontology can draft a workforce plan from that announcement in seconds. When leadership caps how many people can move from existing plants, the plan adjusts itself, splitting each role into people ready now, people who could get there with upskilling, and roles to hire externally. It also shows which existing plants would be stretched by each move.
For critical roles that take years to build, succession recommendations now look across the whole organisation rather than relying on who managers already know. Phenom Behavioral Intelligence adds a data point most succession plans lack: how people work, measured against science-based models for more than 40,000 roles. Assessments are voluntary, used for development, and independently audited for adverse impact.
4. Hiring integrity is now an enterprise security issue
For many European and global companies, candidate fraud is no longer an HR problem; it's an enterprise security issue. Gartner predicts that by 2028, one in four candidates will be fake, and additional research shows organised fake worker schemes have moved into Europe.
A global IT services company with 145,000 employees described the scale. It receives around 14,000 applications a day. AI screening and interviewing have helped it go from a "black hole" to responding to every candidate, with recruiters still making the calls. Because many of its hires work remotely inside client systems, candidate fraud is not just a recruiting risk. It is an enterprise security and client trust issue. The challenge is catching it while staying fair to every genuine candidate.
Agentic candidate fraud detection makes this possible.
The Fraud Detection Agent demo showed how it monitors applicants and immediately flags signals such as a UK address and phone number paired with an IP address in another country. Applicants who cleared those checks moved to an AI interview focused on what each candidate had actually built and deployed.
A coding assessment that mirrors how engineers work today further checked for real skills: a real repository, a coding agent alongside the candidate, and the AI interviewer watching the session. When one candidate submitted her work, the AI interviewer asked why she fixed things the way she did, following her reasoning.
In the final interview, Meeting Intelligence briefed the hiring manager and surfaced an inconsistency between two projects in real time. The Fraud Detection Agent then connected signals no human would catch: a change in answer rhythm, a new connection location, a face that no longer matched earlier rounds, and a device already linked to three other applicants for the same role. One suspicious interview became a connected network.
All of this insight was surfaced to the recruiter and hiring team to make the final call. The team reviewed the evidence and asked the candidate to verify her identity before any decision. The harness also chose which checks apply by role, so an on-site data engineer faces different checks from a fully remote one, with no configuration needed. Before any new signal goes live, it is tested against thousands of simulated candidate journeys to make sure genuine talent is protected.
5. Governance has to be ready before the works council asks
For most European organisations, the hardest part of adopting AI isn't the technology. It's the approvals: works councils, privacy, legal, IT security, and a dedicated AI committee. And as timelines for the EU AI Act keep shifting, it’s critical to take action.
For Phenom customers in the EU using AI across different use cases, our governance framework covers the full development lifecycle, from risk assessments before anything is built to monitoring and audits after it is deployed.
Take the Voice Screening Agent. Employers define their own questions and scoring rubrics, supported by science-based question design. A Responsible Deployment Guide covers reasonable accommodations and candidate consent, and model cards give compliance teams what they need on personal data, third-party components, and fairness. Customers can check for bias against key demographic groups using the same pipeline behind our annual bias audit, which most recently covered 15 million job applications. A person is always in charge, and there is no automated employment decision-making.
Fraud detection raises its own questions, since voiceprints and facial geometry are biometric data. Customers choose whether to use them. We never train on biometric data, and our policy prohibits uses of biometrics that the EU AI Act deems unacceptable.
The same discipline applies to security. We hold ISO 27001, SOC 2 Type II and UK Cyber Essentials certifications, audit our AI agents, and are working towards ISO 42001 for AI governance. Our program is reviewed by five independent auditors and watched around the clock by a dedicated monitoring team, with critical vulnerabilities fixed within seven days.
6. Agents meet people where they work, including on the frontline
To better support hiring managers who often avoid logging into recruiting systems, we're bringing the same agents, ontology, and policies into the places they already work.
Through our partnership with ServiceNow, the Intake Agent starts the conversation inside ServiceNow's AI assistant the moment a role is approved. It writes the job description, recommends sourcing criteria and an interview plan, hands off to the Sourcing Agent to search internal candidates, and invites shortlisted candidates to a voice screen.
From first call to first shift
To support on-the-go frontline workers in Europe, our agents bring hiring, onboarding, and scheduling to their phones in their language. Candidates can apply to jobs in minutes without a CV, and anyone who doesn’t qualify is pointed to other open roles. The Voice Screening Agent calls qualified applicants at convenient times, including evenings and weekends.
After the offer stage, agents line up certifications and safety training so new employees are cleared for day one. On the job, scheduling agents build shifts that follow working-time rules, collective agreements, and certifications, and find qualified coverage in seconds when someone calls in sick. People still make the decisions agents shouldn’t, from who gets hired to who is fit to work.
Results live in Europe
With Voice Screening Agent, candidates reach the interview stage nearly twice as often and two days sooner. In an early careers program running across nine countries, six in ten screens happen outside local business hours when candidates are free. Across deployments, completion rates are above 90% with candidate satisfaction above 4.5/5.
A global biopharmaceutical company also cut its application time from 20 minutes to four, driving a 78% conversion rate. A global provider of government services, which hires around 12,000 people a year for roles from prisons to naval fleets, once scheduled 75,000 interviews a year by hand. Automation now handles it all, and voluntary attrition is down 33% compared with two years ago.
7. HR's place is above the loop
If agents can run more of the work, where do people belong? The biggest value lies in stepping back, designing the flow from end to end, and overseeing how agents and people run it together.
This shift carries real organisational consequences, and there are a few ways to take action. During an executive panel, one leader suggested companies will move from hierarchies toward flows. Another described a new kind of hybrid leadership, with agents appearing in org charts next to people. Accountability for results, quality, and compliance still sits with the leader. Performance management has to move from measuring activity to measuring outcomes, and leaders should continue to guide their teams with empathy.
Their advice for anyone earlier in the journey was simple: move faster than feels comfortable, build AI literacy inside HR early, and partner with digital and finance rather than competing for ownership. As one panellist put it: "Don't try to stop the waves. Start to learn to surf."
8. Measure implementation by outcomes, not go-live dates
Many organisations still aren't seeing tangible value from AI, and the reason is rarely the technology. Outcomes go undefined, data stays siloed across systems, roles and skills aren't updated and works councils are consulted too late. In our Service Innovation keynote, services leaders showed how we are closing that readiness gap.
The first change was how implementation starts. With our Value Acceleration Model, customers walk into a working system already set up for their industry, rather than a spreadsheet of hundreds of questions. Across 32 projects this year, timelines fell from 12 months to as little as 12 weeks, with 43% fewer defects during hypercare.
The AI Bootcamp was also designed to establish clear goals and get a working agent into production in two days, rather than months. Each one starts with a readiness check on data, integrations, workflows, and guardrails, then focuses on a single Hypercell: one specific problem in one industry, geography, workflow, and role.
Our team arrives with a point of view, recommendations, and a preconfigured system. Over two days on site, the team builds, tests, and goes live. Customers leave with a working AI agent, a business case built on their own data, and a 90-day plan.
Staying successful after launch
Support is shifting from reactive tickets to prevention. More than 92% of system alerts are now detected and resolved behind the scenes before customers notice. One global biopharmaceutical company saw 51 potential issues caught early, with 47 resolved automatically and no effort from its team. Its HR technology lead described the difference as moving "from anxiety to confidence."
The bigger opportunity is what happens to people's roles as agents take on transactional work. Recruiters can become talent advisors. Sourcers can become pipeline strategists. HR technology teams can move from maintaining systems to orchestrating AI workflows. Giving people time back doesn't automatically make them more strategic, so new responsibilities, skills and success measures have to come with it.
Preparing for 2027 and beyond
Last year in Munich, the message was that the age of theoretical AI in HR is over. This year in Paris, the conversation moved on again. The question is no longer whether to adopt AI, or even how fast. It is how to make AI precise enough for every role, site, and market to drive the right business outcomes.
See how you can go from first AI conversation to deployment plan in two days. Sign up for your AI Bootcamp.
Get the latest applied AI insights delivered to your inbox.
Sign up to stay up-to-date on the latest original content from Phenom, including new podcast episodes, events, and more.










