Transformation teams must evolve for the agentic enterprise
- Paul Sala
- May 13
- 5 min read
The short version: enterprise AI is moving from assisting people to acting on their behalf. Most transformation teams are not built for that shift. The ones that fail to adapt will deliver technology that the business cannot safely absorb.
The platform direction is clear
SAP, Oracle, Salesforce, Microsoft, ServiceNow, Workday. Different products, different language, same direction.
Every major enterprise platform is moving from systems that record, route and report work toward systems that can recommend, decide, act and escalate inside governed business processes. SAP calls it the autonomous enterprise. Oracle calls it Fusion Agentic Applications. Salesforce talks about the agentic enterprise. The branding differs. The intent does not.
Agents are moving closer to the core of how enterprises operate. The question is no longer whether this is coming. The question is whether your organisation is ready to absorb it.
Most are not. And the reason is rarely the technology.
The real problem is the operating model, not the platform
Here is the pattern we see repeatedly. An organisation invests in an agentic platform capability. The technical implementation goes reasonably well. And then, somewhere between go-live and expected value, things stall.
Adoption is slower than projected. Exceptions pile up in ways nobody planned for. Functional leaders are uncertain about what they now own. Nobody is quite sure who is accountable when the agent gets something wrong. The data the agent is acting on turns out to be less trustworthy than anyone admitted during design.
The technology worked. The operating model was not ready.
This is the gap that traditional transformation teams are not equipped to close. They were built to deliver a system. They were not built to redesign how the business operates once that system starts making decisions.
Most organisations will try to solve this with the team they already have
That instinct is understandable. Building a new kind of transformation team feels like a programme in itself, and most executives are already managing delivery pressure. So the existing team absorbs the new scope, adds a few AI-adjacent roles, and carries on.
The cost of that decision tends to show up later, and in ways that are hard to attribute cleanly. Slower adoption. Qualified rollbacks. Governance gaps that only become visible when something goes wrong. A creeping sense that the value case is not materialising the way it was supposed to.
The issue is structural. A team organised around scope, plan, build, test and deploy is solving a delivery problem. Agentic AI creates a different kind of problem: how do you safely delegate work to something that acts, and how do you govern it after you do?
That requires a different shape of team with a different set of capabilities.
Four shifts that matter
From delivery focus to operating model focus
Traditional transformation teams are measured on whether they shipped the programme. In an agentic context, shipping is the beginning, not the end.
After go-live, accountability for agent behaviour needs to live somewhere. Someone must supervise performance, decide when to intervene, own the improvement backlog and demonstrate that outcomes are actually getting better. A transformation team that cannot set up those structures will hand over a technically functioning system to an organisation that does not know how to run it. That is where most of the value gets lost.
The measure of success shifts from "did we deliver the programme?" to "did the business improve, and can we prove it?" Those are not the same thing, and they need a different kind of team to drive them.
From migration thinking to trust thinking
In a traditional ERP programme, data is treated as a migration workstream. The goal is to move it cleanly. In an agentic environment, that framing is too narrow.
What matters is whether the data is good enough for people and agents to act on with confidence. That is a trust problem, not a migration problem, and it has a different solution.
Agents acting on poor data do not just produce errors. They erode confidence in the whole model. Once users learn to distrust the outputs, adoption collapses in ways that are very hard to recover from. The telemetry might show the agent performing technically. The business reality is that nobody is relying on it.
Data governance needs to move from the edge of the programme to the centre of it, not as a compliance exercise, but as a foundation for safe delegation.
From change management to change capacity management
Communications and training are necessary. They are not sufficient when the nature of work itself is changing.
Agentic AI can affect tasks, roles, decision rights, controls, career paths and management routines all at once. Functional leaders are being asked to move from approving processes to governing outcomes. That is a meaningful shift in what their role actually means, and some will push back. That resistance is often rational, not just cultural.
The version of this change that lands badly sounds like: the agent will do the work now. That creates anxiety, and sometimes legitimate concern. The version that lands better is: you will own the outcome more visibly, and the agent will help execute parts of the work under your governance. The accountability is clearer, not looser.
Understanding that distinction, and designing change programmes around it, is a capability most transformation teams are still building.
From point-in-time testing to operating model testing
Technical testing asks whether the system processes transactions correctly. That bar is too low for agentic environments.
The harder test is whether the operating model holds up. Does the agent stay within policy? Does it escalate at the right moments? Can users actually interrogate its reasoning, or do they just accept the output? Does the process remain functional if the agent is paused?
These are not edge cases. They are the core of what makes delegation safe or unsafe. Getting them right requires risk, legal, compliance and functional leadership inside the testing process, not just reviewing it afterwards. Most transformation programmes are not structured to do that, and the gaps tend to surface at the worst possible time.
The organisational readiness question that nobody wants to answer honestly
Before scaling agentic AI, there is a readiness question that is worth sitting with.
Do you have clear process ownership? Is your master data trusted? Do your functional leaders have the management discipline to govern outcomes rather than just approve transactions? Do you have a value measurement capability that can prove whether anything is actually improving?
If the answer to several of those is no, then the first investment is not in agentic AI. It is in the conditions that make agentic AI safe to deploy at scale.
The transformation team's job is to help the business answer that question honestly, and to sequence the work accordingly. That sometimes means slowing down the technology deployment to accelerate the business readiness work. That conversation requires a different kind of authority and capability than most transformation teams currently have.
What this means for transformation leaders
The transformation team of the autonomous enterprise is not a delivery engine with some extra AI capability bolted on. It is the bridge between business intent, intelligent automation, human accountability and measurable value.
Building that team requires deliberate choices about capability, structure and governance. The specifics depend on the organisation, the platform and the scale of ambition.
But the direction is clear. Transformation teams need to move from delivering technology change to helping the business redesign work, delegate safely, govern intelligently and realise measurable value.
Those are not the same thing. And the gap between them is where most agentic AI programmes will succeed or fail.
Ibex Ascent works with transformation leaders navigating the shift to agentic enterprise. If this is a challenge you are working through, we would welcome the conversation. Visit us at ibexascent.com
Comments