Enterprise AI is forcing a new HR-IT alliance: redesign the work, then deploy the tech
AI is moving from side projects to the core of how companies work. That's exposing a simple truth: adoption stalls if HR and technology leaders don't jointly redesign workflows, roles, and skills.
At Davos, leaders were blunt about the gap. Structures built for traditional software don't fit intelligent systems. If you try to bolt AI onto old processes, you get more friction than value.
Outcome-first, workflow-second, tools last
Earlier digital waves optimized existing processes. AI flips the order. Start with a clear outcome, then decide what the machine does, what the human does, and how they interact.
"HR is not a policy board," said Keith Ferrazzi. "They are work re-engineers." That mindset is the unlock for HR leaders stepping into true product ownership of work.
What joint ownership looks like
Pearson treats workforce design as a shared mandate. CTO Dave Treat put it plainly: "HR leaders need to become product leaders."
CHRO Ali Bebo added, "I like to think of Dave as the tech architect, and we are the capability architect… that's designing the systems and the processes and how that comes together when it interfaces with humans."
Decompose work before you automate it
IBM broke the problem apart at scale. "We decomposed our company into 490 workflows," said Mohamad Ali, who leads IBM Consulting. One of those was HR itself.
The lesson: AI-driven work cuts across departments, so it requires executive sponsorship and cross-functional governance. "You have to start at the top," Ali said. "These workflows cut across."
The skills language problem
Most orgs can't describe work at the task level, which makes it hard to match people and AI. Treat's call-out: "You need to understand the data at a task, job role level. We have a long way to go even in mapping humans to work in a traditional context."
Without a shared skills taxonomy tied to tasks, companies overgeneralize roles and underutilize both people and models.
Pixelisation of work
Some AI-first teams start by asking, "What can the machine do now?" Then they assign the human to the parts that matter most: context, judgment, relationship, and exceptions.
Ferrazzi called it the "pixelisation of work"-breaking processes into smaller components so you can design the human-machine split with precision.
Learning that teaches, not just answers
Bebo's stance on enablement is clear: "We don't want them to give the answers. We want them to teach."
That means embedding learning into the flow of work, using AI as a coach that explains steps, surfaces next-best-actions, and improves proficiency in real time.
Avoid the automation trap
"Companies that are focused on AI just for automation's sake are hitting a wall," Treat said. The bigger win is productivity, quality, and growth-not just fewer clicks.
Automation is a byproduct. The product is redesigned work that compounds value.
Your CHRO playbook: 12 moves to ship AI-enabled work
- Co-own a Work Design Council with the CTO. Shared budget, shared outcomes, shared backlog.
- Pick 3-5 cross-functional workflows that matter (offer-to-accept, ticket-to-resolution, claim-to-pay). Map tasks end to end.
- Define the human-digital RACI per task: what the model does, what the human decides, and escalation paths.
- Stand up a skills taxonomy at task level. Tie it to proficiency rubrics and career paths.
- Instrument everything. Track cycle time, error rates, rework, and time-to-proficiency.
- Shift from training events to "learning in the loop." Micro-coaching, prompt guides, and in-product tips.
- Set policy and guardrails: data access, privacy, bias checks, and human-in-the-loop for sensitive decisions.
- Create an AI prompt/design system. Standardize patterns for approvals, summarization, and decision support.
- Update rewards. Recognize skill growth, model stewardship, and workflow outcomes-not just output volume.
- Launch 6-8 week pilot sprints. Ship small, measure, iterate, scale.
- Stand up change comms that show before/after videos of work, not decks.
- Plan redeployment, not reductions. Match people to higher-value tasks as automation lifts low-value work.
Metrics that tell you if it's working
- Cycle time and cost-to-serve per workflow
- Time-to-proficiency for re-skilled roles
- Error rate, rework rate, and compliance exceptions
- Employee NPS and tool adoption rates
- Workflow coverage by AI (tasks automated or co-piloted)
- Model quality drift and escalation frequency
Risks to manage early
- Decision opacity: log prompts, outputs, and human approvals
- Bias and fairness: test on edge cases, monitor outcomes across groups
- Data leakage: clear data boundaries and anonymization where needed
- Change fatigue: rotate pilots, limit concurrent changes per team
What this means for HR
You're moving from policy to product. From job descriptions to task design. From training events to continuous, AI-assisted skill building.
The companies pulling ahead aren't waiting on perfect tools. They're rebuilding work, together, at the workflow level-and letting AI slot into the design, not dictate it.
If you want a quick primer on how different roles can skill up for AI-enabled workflows, explore curated learning paths by job at Complete AI Training.
For broader context on AI's economic and workforce impact, see the topic hub at the World Economic Forum.
Your membership also unlocks:
