Part of the system, not a plugin
She holds the same modelling and analysis tools you do: dictionaries, tables, links, forecasts, dashboards. Not a window onto the model. A hand inside it.
CLARA AI
CLARA
The analyst who has read every model you own.
The FP&A agent built into Novi: building, analysing, narrating and remembering alongside you, with every dimension, link and formula in plain view. Not a chatbot bolted to the side. The model's own intelligence.
01· Born native
Born with AI
Born with AI. Not retrofitted for it.
Every platform has an agent now; that part isn't special. What matters is what the agent stands on. Most were built before AI was practical, then bolted one on top of a cube it can't really read. Novi was built when AI was already real, so the modelling layer was designed from the start to be read, traversed, and reasoned over by an agent. CLARA doesn't query your model from outside. She lives inside it. Every dimension, every link, every declared relationship is directly visible to her, not as data to be fetched, but as semantic structure to be understood.
Where there's no real model, there's nothing for AI to reason over. Novi gives the model. CLARA gives the reasoning.
02· Why she's different
The substrate
Why she can do what bolted-on agents can't.
The difference isn't the model behind CLARA. Capable agents are everywhere now. The difference is what she stands on, and what that lets her know.
She knows all your models
No one person holds the whole global model in their head. CLARA does, every corner of every model. In a large organisation, that institutional memory is the thing no human can offer.
She knows your formulas, so she knows your business
Every formula you write is a statement about how your business works. Because CLARA reads them, she speaks your semantics, not generic finance.
03· Where she shows up
The whole lifecycle
She's there at every step, not just at the end.
CLARA accompanies the work from the first dimension to the board read, and keeps the reasoning as she goes. FP&A, IFRS and controlling knowledge are on hand throughout.
Build
Spin up a dimension, draft a table, wire a link, scaffold a forecast, from a sentence, into a structure you can check.
Analyse
Walk a variance to its driver, stress a scenario, audit a node for breaks, with the model's full structure in view.
Narrate
Draft the board commentary, build the dashboard, turn the model into a Story, in your voice, not boilerplate.
Remember
Collect the notes, decisions and reasons as you work, so the model carries why, not just what.
04· Answers in structure
Not a wall of text
She answers in structure, not paragraphs.
Ask CLARA a question and the reply isn't prose to copy out by hand. It's the thing itself: a hierarchy, a ranked output, a report view, a chart, a dashboard you can open and keep working in. The answer is a live model object, not a screenshot of one.
05· How she speaks
Her voice
Direct. Proactive. Precise.
CLARA speaks the way a good analyst does: she notices, she proposes, she gets to the point. She never pads an answer with hedging, and she never breaks character with AI disclaimers.
"Revenue is down 4.2% vs prior month, primarily DACH. Shall I draft the variance commentary?"
"That formula double-counts intra-group revenue. Want me to net it out before it reaches the consolidation?"
"Three scenarios still reference last quarter's headcount table. I can repoint them in one pass."
Direct · Proactive · Precise · Never breaks character with AI disclaimers
06· The contract
Built for an educated modeller
Built for someone who learned to model.
CLARA is a partner, not a vending machine. "Go build me a model" isn't a prompt. It's a wish. The fundamentals don't disappear because there's an agent in the room; they're what let you direct one. To ask the right question, and to judge whether the answer is right, you have to understand the modelling underneath. That's the honest part most tools won't say out loud: the magic button finance has been sold for a decade doesn't exist, and pretending it does is how models go quietly wrong.
Doesn't work
"Go create a model."
A wish handed to a tool, with no structure behind it. You can't check what comes back, because you never said what you meant.
Works
"Spin up a DACH revenue dimension on the SaaS template, monthly, mirroring the EMEA build."
Direction from someone who knows what they're asking for, and can tell at a glance whether CLARA got it right.
So we teach the craft first. Learn the modelling principles, then CLARA takes the wheel. Start at the Learning Hub →
07· The model learns you
Whose knowledge it is
The model learns you. The knowledge stays yours.
Every modelling decision you make (name a dimension, declare a link, write a variance rule, define a hierarchy) is part of the structure CLARA reads. Over months and years the model carries a complete account of how your organisation thinks about itself: not just what the numbers are, but how this business reasons. No one person in finance sees every decision. CLARA does.
The institutional understanding that accumulates belongs to your organisation, not to Novi.
It accumulates inside your own model. Read it, export it, take it with you. Like everything in Novi, it is yours.
She doesn't replace the finance professional. She removes everything that was stopping them from being one.
Meet CLARA where the work happens.
CLARA ships with the platform on September 1, 2026. Early-adopter seats are open now. Tell us about your team and we'll get you in.