Most enterprises are trying to add AI to existing work. The ones creating real value are doing something fundamentally different: they're redesigning the work itself.
What You Need to Know
- "Automation" is the wrong mental model for enterprise AI. Automation means doing the same thing faster. Redesign means doing a different, better thing.
- When AI handles mechanical work, it doesn't just save time - it changes the nature of the remaining work. Roles shift from processing to judgement, from data gathering to decision-making, from routine to exception-handling.
- The most successful AI deployments we've seen involve process redesign before AI deployment. The companies that redesign first get 3-5× more value than those that automate in place.
- Change management for work redesign is different from change management for new tools. You're changing job descriptions, not just workflows.
- Adoption succeeds when people feel their work is becoming more interesting and more valuable, not when they feel they're being monitored or replaced.
70%
of AI-driven change programmes fail due to poor change management
Source: McKinsey & Company, The State of AI in 2022, December 2022
The Redesign Spectrum
Enterprise AI interventions fall on a spectrum from "bolt-on" to "redesign":
Level 1: Bolt-On. Add an AI tool alongside existing workflows. Users can use it or ignore it. Minimal disruption, minimal value. Example: ChatGPT available on desktops for "anyone who wants to use it."
Level 2: Assisted. AI pre-processes inputs or suggests outputs within existing workflows. The workflow is the same; the AI speeds up specific steps. Moderate value. Example: AI extracts data from claim forms, but the adjudicator still follows the same review process.
Level 3: Augmented. AI handles routine cases end-to-end. Humans focus on exceptions, quality review, and complex decisions. The workflow changes, and roles shift from processing to judgement. High value. Example: 80% of straightforward claims are processed automatically; the team handles the complex 20%.
Level 4: Redesigned. The entire process is redesigned with AI as a core component. New roles, new workflows, new metrics. Very high value. Example: The claims team becomes a "claims intelligence" team: part quality assurance, part data analysis, part continuous improvement. AI handles processing; the team handles strategy.
Most enterprises are at Level 1 or 2. The value lives at Level 3 and 4.
How Roles Change
When AI handles the mechanical 80%, what do people actually do? This is the question that creates anxiety, and the one that needs clear, honest answers.
What people stop doing:
- Manual data extraction from documents
- Routine information retrieval
- Standard-pattern processing that follows clear rules
- Report generation from structured data
What people start doing:
- Reviewing AI outputs and making judgement calls on edge cases
- Handling complex, ambiguous situations that require experience and context
- Continuous improvement: identifying where the AI struggles and training it
- Relationship work: client communication, stakeholder management, complex problem-solving
This is genuinely better work. More interesting, more valuable, more aligned with why people chose their profession. A claims adjudicator who wanted to evaluate complex claims (not copy data from forms into spreadsheets) finds the AI-redesigned role more satisfying.
But the transition isn't automatic. It requires intentional support.
Making the Transition Work
1. Involve the Team in Redesign
The people doing the work should help design the new way of working. They know the edge cases. They know what's genuinely complex vs what just seems complex. They know which parts of their job they'd happily give to an AI and which parts they'd fight to keep.
2. Communicate Honestly
"AI won't replace you" is only credible if it's true, and if the new role is clearly defined. Be specific: "Your role will shift from processing 40 claims a day to reviewing the 8-10 that the AI flags as complex, and improving the AI's handling of the other 30."
3. Invest in New Skills
The redesigned role requires new capabilities: AI output review, data quality assessment, exception handling at higher complexity. Budget for training, and make it practical (working with the actual system), not theoretical (slides about "AI literacy").
4. Measure Differently
Old metrics (claims processed per person) don't apply to the new model. New metrics: quality of exception handling, AI accuracy improvement rate, client satisfaction, time to complex resolution.
- Won't some roles be eliminated entirely?
- In some cases, headcount requirements will decrease for specific functions. But in our experience, AI-driven redesign more often changes the composition of teams than eliminates them. The reduction in routine processing is typically offset by new needs: AI oversight, quality assurance, continuous improvement, and handling the increased volume that faster processing enables. Be honest about this with your teams, and plan for redeployment and reskilling where genuine displacement occurs.
- How long does a work redesign take?
- The redesign itself (mapping the new process, defining new roles, designing the transition) takes 2-4 weeks. The transition (training, phased rollout, stabilisation) takes 2-3 months. Budget for the transition time; it's where the value is captured or lost.