Why transformation work needs an Agentic Transformation OS
AtlanticM&A7 min read
An Agentic Transformation OS is software in which a transformation programme's work is actually done, with agents that read the whole project and draft or update the plan, the briefings and the decks as that work changes. AtlanticM&A is one. This post sets out why I think the category had to exist.
Where the work really happens
Watch a workstream lead on a typical programme for a week. The thinking happens in a meeting. The planning happens in Excel. The steering deck happens in PowerPoint. Then, usually late on a Thursday, someone re-keys the outcome into the programme tool so it can report on what already happened.
The tool is accurate for about a day. The real work carries on around it.
That is not a failing of the people. Integration, divestiture, transformation and cloud migration share a structure that no software category was built for. The method is institutional: academic research and practitioner common sense, packaged by the large consulting firms over decades and rented back to clients one deal at a time. The work is bound to a single project, with its own stakeholders, services, sequence and timeline. And the volume comes in bursts: a corporate development team might run three deals one year and one the next.
So practitioners have stitched it together from a generic project tracker for the tasks, plus method and labour rented from advisory firms, one engagement at a time.
A system of record and a system of work
A system of record is where you file things afterwards. A system of work is where the job is done.
In AtlanticM&A the charter is written in the platform. The work plan is built there. Workstream leads record delivery as it happens, so the status report and the steering deck come out of the work instead of being re-keyed from it.
A system of record has usually needed a person whose job is to keep it fed: chasing the updates, keying them in, and pulling the report back out when the steering committee asks. AtlanticM&A does not need that role. Meeting transcripts, documents and imported plans become project context through automation and AI, and changes to the plan arrive as proposed updates for the team to accept. Getting the answer back out is a question to the project assistant or a pack from Report Centre.
The Agentic Transformation OS page draws this as five rings, each resting on the one beneath:
| Ring | What it does |
|---|---|
| Record | Store |
| Structure | Relate |
| Work | Do |
| Agents | Execute |
| Assurance | Attest |
Most software stops at the second ring: it stores the documents and relates the tasks. The third ring is the one that matters, because an agent with no real work beneath it "is producing plausible material about work happening somewhere else."
Why a chat panel is not enough
A lot of project software now has a chat panel. You type a question and get an answer drawn from the project data. That is useful, and it is a different thing.
A chat panel is an interface onto data that was structured for generic project management. It fires when prompted, produces text and waits. In AtlanticM&A the data model is built for the work from the start: work plans, the stakeholder graph, TSA services, synergy claims, the Wave Plan and RAID, all related to each other. Agents read that state as it changes and act on it.
They produce artefacts: briefings, decks, addendum drafts, sequenced plans. They also propose changes to that state: a stakeholder's position or a new action from a meeting, a moved date in the work plan. Each one waits for a person to accept it. The user is not the engine.
One concrete example is Virtual Walk the Walls. It works out the critical path to TSA exit from the dependencies between workstreams, so slippage is visible before it cascades, without a hotel ballroom full of sticky notes and string.
Follow the line down
What arrives on a programme (the brief, documents, meeting transcripts, email, data) becomes linked project records: a risk tied to the milestone it threatens, a synergy tied to the stakeholder who owns it. Named agents work across those links. One drafts briefings, one maps dependencies, one builds decks and one runs the meeting cadence.
So when the steering pack says a synergy is at risk, you can follow the line down to the RAID entry and, where a meeting raised it, to that meeting. The project assistant works the same way: it reads the whole project and says where each fact came from. That is the fifth ring in practice.
What it looks like from a standing start
The film below runs two minutes fifteen seconds. A half-page brief becomes strategic direction, charters, a work plan and a roadmap, in about four minutes of build time, with Walk the Walls marking the critical path afterwards.
That is the agents ring acting on the work ring. The build writes into the same charters and work plans the team then runs, not into a separate document that has to be copied across. AI Power Build has its own page.
Who keeps the capability
McKinsey's September article puts a number on what agents can do for a transformation office: time spent on many core tasks can be reduced by 35 to 40 percent, and in some cases by 70 percent or more. I think that is right, and it is why the centre of a programme can now be much smaller.
That matters for what consulting costs. It has mostly been priced on time: people paid for what their hours are worth, not a share of the equity or the outcome. The premium in the day rate paid for the pyramid, because a large, trained team was hard to assemble reliably at short notice, and few clients could do it themselves. With the pyramid reduced or gone, the reason for the premium goes too, and the fee should come back to the cost of delivery.
The large firms have an answer to that. In McKinsey's design the agents run on its own platforms, Wave and Lilli, and a proprietary capability rented on outcomes ties the price to the value of the work rather than to what it costs to deliver. The client is more efficient, but it still pays a premium, and it does not fully own the capability.
We take the opposite view. A smaller centre should leave the company in control of its data, its processes and the capability to run the next programme, not hostage to a consulting firm's proprietary AI. AtlanticM&A is priced per project, and what the team builds in it (the charter, the plan, the template for the next deal) stays with the company. For any company that buys from the large firms, that brings the fee back to the cost of delivery now, not eventually.
Where the judgement stays
The fifth ring is the one I care about most. Proposed against accepted. Source attribution. An audit trail on every change the AI made. The number someone signs.
The platform presents; people decide. It does not score deals or recommend whether to proceed. Post-deal work is mostly execution: the major moves are clear from the deal thesis, and the detailed plan emerges as data arrives from the programme itself. What an Agentic Transformation OS adds is the operating layer that carries those moves out, and augments the team doing them.
The full argument is on the Agentic Transformation OS page.
About AtlanticM&A
AtlanticM&A is the Agentic Transformation OS. Its agents work inside the programme, both augmenting and automating the work of the transformation team: work plans, governance, stakeholders and reporting in one place, with a live value bridge that moves as the work does. It is a system of work rather than only being a system of record: the charter is written in it and the plan is built in it. Meeting notes come back as proposed updates to the plan and the RAID log, the team decides what stands, and every change the AI makes carries an audit trail. The platform does the slide-building, status reporting and stakeholder briefings that usually keep a consulting pyramid busy, so a small senior team can run the programme more nimbly and for less.