Forecasting with AI: model, judgment, or both?

Forecasting is where the CRO's credibility lives. AI can help — but replacing manager judgment with a model is rarely the right first step.

The starting point

Gartner's sales operations survey found only 45% of sales leaders and sellers had high confidence in their organization's forecasting accuracy, and only 47% believed their organization had high-quality data. Low confidence leads to decisions on intuition: unnecessary incentives, wrong hiring pace, discounts approved to "make the number".

Three ways to forecast

Approach	Strength	Weakness
Manager roll-up (judgment)	Knows context the data misses: a champion left, a budget freeze	Optimism, sandbagging, inconsistent across managers
AI / statistical model	Consistent, uses all history, no politics	Blind to new information not in the data; weak when the market shifts
Combined	Model as a baseline; judgment adjusts with stated reasons	Needs discipline to record and review adjustments

How to introduce AI forecasting

Run in parallel: record the model's number and the roll-up at the same point each period.
Score both: after each period, compare each with actuals by segment.
Require reasons: when a manager overrides the model by more than a set amount, they write why.
Adjust weights: after 2–3 quarters, rely more on whichever is more accurate in each segment.

What a model needs

Consistent stage definitions, honest close dates and enough history. If your stages mean different things in different teams, fix that first — or the model will learn the inconsistency.

Telling the board

Report a range, not just a point, and say how the forecast is made. Gartner's 2026 survey found 60% of CSOs believe their revenue number is largely driven by elements outside their control — a reason to be precise about what you can and can't predict.

What this means for the CRO

Don't let AI forecasting become a black box you have to defend. The combined approach — model as baseline, managers adjust with reasons, both scored against actuals — gives you a better number and a better story for the CFO and board.

Personally review the override log each quarter. It tells you which managers know their pipeline and which are guessing, and it is one of the best coaching tools you have.

Agree with your CFO how the combined forecast will be presented — a range with the method stated — so that a miss is a discussion about assumptions rather than about credibility.

Diagram

Forecasting with AI: model, judgment, or both?

Real example

Gartner's State of Sales Operations survey (February 2020) found that only 45% of sales leaders and sellers had high confidence in their organization's forecasting accuracy, and only 47% believed their organization had high-quality data. Source: Gartner — Less than 50% of sales leaders and sellers have high confidence in forecasting accuracy (Feb 12, 2020) ↗

Use it

Quick check · 3 questions

1.What is the recommended first step when introducing an AI forecast?
2.What is the main weakness of a pure model forecast?
3.Why ask managers to write a reason when they override the model by a lot?

0 of 3 answered

Sources

Last reviewed 2026-10-01.