New

Peaky raised €1.1M, led by JK Invest

→

All articles

Company

/

Written by Peaky

Continuous driver-based forecasting, built and maintained by an agent.

Driver-based forecast (€K)
PlanActual
The forecast moves when the drivers move, not when someone rebuilds it.
Updated today
Driver-based forecast (€K)
PlanActual
Updated today

The forecasting agent is the Peaky agent that builds and maintains your forecast.

Most finance teams treat the forecast as a quarterly artifact. You set the drivers, you run the math, you produce the file. Three months later you do it again. Between refreshes, the forecast either drifts quietly or gets manually patched in a series of unloved spreadsheet updates.

Peaky’s forecasting agent is built around a different assumption: that the forecast should be current the moment you open it.

What the forecasting agent does

It does three jobs.

It builds the initial driver-based forecast. You describe your business in plain English. The agent proposes a driver structure, the formulas behind each driver, and the initial assumptions. You review and adjust. Most teams have a working forecast by lunch on their first day.

It maintains the model continuously. Every time the underlying data refreshes, the agent recomputes the model. New revenue in Stripe flows through to the forecast automatically. New hires in your HR system update the burn. Pipeline movement in your CRM updates the conversion math. There is no batch refresh because there is no batch.

It responds to plain-language requests. You can ask it to add a new driver, run a scenario, or adjust a formula. It proposes the change. You approve or refine. The job that used to require finding the right cell in a 200-tab spreadsheet now happens in a sentence.

The work the forecasting agent takes over

The work it takes over is the work that most finance teams spend the largest single block of time on: rebuilding the model after the close.

This is the work that historically meant pulling exports, reconciling actuals against forecast, finding the formulas that broke, replacing hardcoded values, validating that the new totals reconciled, and producing a version that the rest of the company could rely on.

The agent does this continuously. There is no “the model is broken” moment because the model has been updating against reality the whole time.

The downstream effect is that the variance column in the monthly board pack is no longer a surprise. The team has been watching the variances accumulate in real time, in a system that has already updated the forecast in response.

How forecasting fits with the other agents

The forecasting agent does not work alone.

The ledger-mapping agent maps your ledgers so the forecast is built on clean, structured books. Without it, the forecasting agent would be working with the raw accounting export, which is too messy to model directly.

The reporting agent turns the live forecast into the three statements: the P&L, balance sheet, and cash flow.

The drift agent watches the assumptions behind the forecast and surfaces drift when they no longer match reality. The forecasting agent can adjust based on those signals, but the drift agent is the one watching.

The budgeting agent runs the budgeting flow, letting department heads update their own lines without breaking the master model.

The agents are designed to specialise. The forecasting agent is the modeller. The rest of the system feeds it, validates it, and turns its output into the deliverables the rest of the company needs.

A worked example

Imagine your sales team is planning to hire two new AEs next quarter. In a traditional FP&A workflow, you would update the headcount assumption in your spreadsheet, hope the formulas all reference that cell correctly, and rebuild the downstream impact manually.

With Peaky, the conversation looks like this. You tell the forecasting agent you want to model a scenario where two new AEs start in Q3, with 90-day ramp and historical pipeline conversion rates. It proposes the driver structure, including a 90-day ramp curve, the cost impact on burn, and the pipeline impact on net new MRR. You review. You adjust the ramp assumption based on what your team actually sees. The agent updates the scenario.

Total time: about ten minutes. The scenario lives next to your base plan. When pipeline conversion rates drift in your CRM, the scenario updates with them. When the new hires actually start, the model knows.

This is the difference between a forecast you maintain and a forecast that maintains itself.

What you experience

Three things change when an agent is doing the modelling work.

The first is time. The four days a month of model rebuilding go away. They are replaced by short conversations with the agent when you want to add or adjust something. The strategic work the finance team got into the job to do becomes available.

The second is accuracy. The model is always reconciled against the latest actuals. There is no version of the model that is “the right one” because every version is current.

The third is confidence. The forecast you look at is the forecast that just updated based on the latest data. The decisions you make on top of it are decisions made on what is true today, not what was true last month.

What the forecasting agent cannot do

It is not a strategist. It does not tell you what to forecast. It does not tell you what the business should look like.

It builds the model you describe, refreshes it continuously, and proposes adjustments when the math is straightforward. It is excellent at the mechanical work of forecasting. It is not pretending to do the judgement work.

The judgement still lives with the finance team. Peaky makes it possible to spend more of your time on it.

See continuous forecasting in action

You can try it by starting a Peaky trial at peaky.ai. Most teams have their forecast running on real data within an hour of signup.

The financial infrastructure for scaling without surprises.

Previous

Why we founded Peaky.

Next

How to spot assumption drift early.

Keep exploring

Getting started

From your Excel plan to a live plan in a day.

Integrations

The systems Peaky connects to, and what each one brings in.

The Forecast Audit

Nine questions to find out whether your forecast is a report or a tool.

Peaky app icon

Sync or sink.

Ask AI for a summary of this website