Forecasting: AI model, manager judgment, or both?
When it comes up
When forecast misses hurt credibility with the CFO or board, when a forecasting tool is proposed, or when managers' calls diverge widely.
Your options
Manager roll-up
Fits when: Little clean history; small number of large deals; rapidly changing market.
Watch out: Optimism and sandbagging; inconsistent across managers.
AI / statistical model
Fits when: Many deals, consistent stage definitions, several quarters of clean history.
Watch out: Blind to information not in the data; weak after market shifts; hard to explain.
Combined: model baseline, recorded human adjustments
Fits when: Most organizations once data is reasonable.
Watch out: Needs discipline to record override reasons and score both methods.
How to decide
- Run the model in parallel with the roll-up and record both at the same point each period.
- Score both against actuals by segment.
- Require a written reason for large overrides.
- After 2–3 quarters, weight the method that is more accurate per segment.
- Report a range and the method to the CFO and board.
Trade-offs
Models bring consistency and remove politics but miss context; judgment brings context but adds bias; combining them needs process discipline but usually gives the most defensible number.
Real examples
- Gartner's State of Sales Operations survey (February 2020) 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. Source: Gartner — Less than 50% of sales leaders and sellers have high confidence in forecasting accuracy (Feb 12, 2020) ↗
- Gartner's July 2026 survey of 210 chief sales officers found 60% believe their revenue number is largely driven by elements outside of their control. Source: Gartner — AI agents will outnumber sellers 10 to 1 by 2028, yet fewer than 40% of sellers will say agents improved productivity (Jul 28, 2026) ↗
What to check first
- Whether stage definitions and close-date rules are written and applied consistently.
- How much clean opportunity history you have by segment.
- Your current forecast error by segment, to have a baseline.
Learn and use
Sources
- Gartner — Less than 50% of sales leaders and sellers have high confidence in forecasting accuracy (Feb 12, 2020) ↗
- Gartner — AI agents will outnumber sellers 10 to 1 by 2028, yet fewer than 40% of sellers will say agents improved productivity (Jul 28, 2026) ↗
Last reviewed 2026-10-01.