Agentic AI won't move the needle without early change leadership
If you want AI agents to create real business impact, start change management on day one and shape it to fit each employee segment. Delay it, and you invite two problems: early adopters spin up rogue tools, while others dig in and resist because they fear job loss.
Recent reports show the gap. Many organizations are seeing little or no return from AI, and only a small share have a formal change strategy in place. As Michael Connell, COO of Enthought, advises, treat change management like a first-class budget line and bring end users into agile cycles early and often.
AI moves fast; your change program must run in parallel
Rolling out AI agents isn't linear. Models and platforms shift every quarter, making yesterday's stretch goals feasible today. The hard parts haven't changed: aligning on strategy, building strong governance, deciding where to experiment, and moving from pilots to production.
Start with shared language. "Agent" can mean different things across teams. Most agents today support a single function and aren't fully autonomous. Set expectations now, because their sophistication will increase.
Segment the workforce by AI responsibility
Map change programs to how people will actually participate in AI work. Five segments usually cover it:
- Executives: Align strategy, outcomes, and investment priorities.
- Compliance leaders: Risk, security, and data governance for the AI guardrails.
- Subject matter experts: Provide domain logic, validate agent accuracy, and tune performance.
- End users: Use agents in daily workflows where value extends beyond productivity.
- Innovators: Cross-functional teams that experiment, evaluate vendors, and build proprietary agents.
Guide executives to two or three strategic bets
With AI buzz everywhere, every department wants in. Spreading resources across a long wish list yields shallow results. Brandon Sammut, chief people officer at Zapier, recommends centering the AI agenda on two to three opportunities tied to existing priorities, with a clear "why" and "why now."
Recommendation: Build executive alignment as a core competency. Assign ambassadors to meet with leaders, listen for force multipliers, and draft crisp vision statements. Then communicate relentlessly to secure buy-in on a focused portfolio.
Co-create governance before you experiment
Innovation usually outruns controls. That's risky with reasoning agents that can bypass traditional safeguards. Kamal Anand, president and COO of Trustwise, flags the missing pieces: embedded trust frameworks, real-time governance, energy-aware infrastructure, and people who understand dynamic agent behavior.
Elad Schulman, CEO and co-founder of Lasso Security, adds a practical boundary: define which tasks agents can do alone and where human oversight is mandatory, especially around sensitive data and critical operations.
Recommendation: Convene compliance, security, and data leaders to prioritize guardrails by material risk. Publish permitted tools, approved data uses, oversight rules, and escalation paths. For additional structure, reference the NIST AI Risk Management Framework (NIST AI RMF).
Turn subject matter experts into stewards
If processes depend on tribal knowledge, scattered data, or manual exceptions, agents stall. Boobesh Ramadurai of LatentView suggests codifying business logic, standardizing metadata, setting escalation rules, and connecting systems so agents can act with context. Shift analysts toward orchestration and build live feedback loops.
Dave Killeen of Pendo notes that agent outputs won't be perfectly deterministic. Teams must see how agents behave in real workflows, detect drift, and know when and how to intervene.
Recommendation: With HR, define objectives and incentives for experts who document logic, validate outputs, and improve prompts, policies, and workflows. Make SME stewardship a recognized part of performance plans.
Reduce fear for end users with real learning paths
Cindi Howson of ThoughtSpot highlights the mood on the ground: many workers worry about replacement, and AI literacy is low. Geoffrey Godet, CEO of Quadient, reframes it: AI replaces tasks first, which frees leaders to redesign roles for higher-value work.
Training matters, but it's bigger than tool skills. People need better questions, prompting techniques, an eye for hallucinations, and stronger critical thinking.
Recommendation: Create ongoing learning that isn't limited to job transitions. Offer AI literacy, role-based labs, and office hours tied to active pilots. For structured upskilling by role, see curated paths at Complete AI Training.
Aim innovators at customer-facing wins
Most SaaS platforms are adding agents into employee workflows, which is useful. The next leap is agents inside customer experiences: patient support in healthcare, guided investing in financial services, smarter shopping in retail.
Ashley Moser, CCO at MelodyArc, advises bringing frontline teams into experiments early. They already know customer pain points and can shape where agents matter most.
Recommendation: Treat these builds like product work. Pair innovation teams with frontline operators, product managers, designers, and risk partners. Set explicit success metrics that balance speed, quality, customer satisfaction, and safety.
A 90-day action plan
- Week 1-2: Define common AI terminology. Segment the workforce by AI responsibility. Publish a one-page "why" and "why now."
- Week 2-4: Align executives on two or three bets with outcomes, owners, and funding. Stand up an AI governance council and agree on top risks and guardrails.
- Week 4-6: Select two pilot use cases (one internal, one customer-facing). Nominate SMEs and end-user champions. Set review and escalation rules.
- Week 6-10: Build feedback loops, test for drift, run sandboxed adversarial tests, and log decisions. Launch role-based learning and open office hours.
- Week 10-12: Decide go/no-go for production. If "go," define supervised autonomy levels, metrics, and runbooks. If "no-go," document lessons and redeploy talent fast.
The bottom line
Agentic AI is a people change before it's a tech rollout. Start change leadership early, segment by responsibility, and make governance, learning, and SME stewardship part of the operating model. Do this, and adoption becomes the final, critical mile that turns AI into outcomes.
Your membership also unlocks:
