Table of Contents
Key takeaways
- Most AI planning pilots work, then die: they're rarely rebuilt because they ran on disconnected spreadsheets, not a shared source of truth.
- Pigment's fix is a three-step framework: unify your systems into one live model, build on that model in plain language, then automate the analysis that sits on top of it.
- The Modeler Agent, launched in March 2026, now creates 70% of new metrics and formulas built in Pigment.
- Unilever's North America overhead application was built and launched in 25 days with the Modeler Agent, which could have otherwise taken four to six months.
- Pigment Frames turns the interface layer itself into something you describe rather than build, while staying connected to governed, live data.
- Pigment is now offering a free trial environment, a first for an enterprise planning vendor.
A pilot that worked, then disappeared
Pierre Thompson, Pigment's AVP of sales for Central Europe, opened the afternoon with a pattern he said he's heard in nearly identical form in Frankfurt, Zurich, Vienna, and Warsaw. A team runs an AI project. It works. Someone builds in a weekend what used to take a full quarter. Then it stops, not because it failed, but because nobody could repeat it.
Thompson's diagnosis: the problem was never the model's intelligence. It was the foundation underneath it, usually a CSV export, a spreadsheet three people maintained three different ways, or a screenshot pasted into a prompt. The logic lived in one person's head, and the next month someone started over from zero.
Pigment's answer is a single live model that finance, sales, HR, and supply chain all plan against, with AI layered on top of it rather than bolted onto exports. Thompson framed the path there in three steps.
Unify: one model instead of 10 versions of the truth
Connecting an AI tool to a CRM or an ERP is the easy part, and most teams in the room had already tried it. The harder problem is that a CRM and an ERP define the same metric differently, so connecting AI to 10 systems returns 10 versions of the truth. Pigment's approach is to unify those systems into one live model so that when sales changes a forecast, finance sees the margin impact immediately and HR sees the hiring implications in real time, on the same numbers, with the same access rules, for every person and every agent.
Build: intent modeling instead of static prototypes
The second step is what Pigment calls intent modeling. A planner describes what they need in plain language, or hands the Modeler Agent a requirements catalog, an Excel file, or a slide deck, and the agent builds the model, formulas, and dashboard directly.
The adoption numbers are concrete:
- The Modeler Agent shipped in March 2026, seven months before Catalyst Berlin.
- It now generates 70% of all new metrics and formulas created in Pigment.
- Unilever's North America overhead application, previously estimated at four to six months, launched in 25 days, a sevenfold speedup the team called its fastest turnaround yet.
Build, part two: Frames makes the interface generative
A finished model still has to reach an executive audience, and exporting it has traditionally meant choosing between a custom interface or live, governed numbers. Pigment Frames removes that trade-off. You describe the view you want, and an agent builds it on top of the live model, using the same data, calculations, and access rights as everything else in the platform.
Two examples from the Catalyst Berlin session:
- Palo Alto Networks had strong models in Pigment but was still rebuilding its executive presentation by hand every cycle. It now builds that view directly on the live model.
- Supercell had started building its own interfaces with external AI tools, liked the result, but saw the risk of running the business on something disconnected from the real model. After bringing the same approach into Pigment, the team said, "Our only limit is our imagination."
Automate: an analyst that works while you sleep
The third step is automation. Pigment's Analyst Agent monitors the business continuously, flags what changed, and delivers a decision-ready report overnight, fully traceable to the underlying numbers.
A company's plan might sit in Pigment while its goals live in a meeting transcript in Notion. Agent Connectors let the Analyst Agent read that external context and tie it directly to the planning cycle. Notion is the first integration; Google Drive and Slack are coming soon.
On governance, Pigment's answer to "where does the data go" is straightforward: an agent inherits the access rights of the person using it, can't see what that person can't see, and every action it takes is logged and traceable.
What this means if you're evaluating AI for planning
Julien, who leads Anthropic's European business and is a Pigment customer, put it this way: "Pigment is showing what's possible when AI becomes a true enabler in strategic planning."
For teams weighing where to start, the practical sequence from Catalyst Berlin is: get your systems into one model before you add AI on top, use the Modeler Agent to build rather than write a requirements document first, and only then automate the reporting layer. Pigment is now offering a free trial environment to test that sequence directly, a first among enterprise planning vendors.
Request a demo to see it on your own data.
