What is thesis drift and how do you prevent it?

Last reviewed: 1 September 2026

Thesis drift is when the deals a fund actually does diverge from the strategy it committed to LPs. Reuben AI prevents it by scoring every live deal against the encoded mandate and surfacing drift before IC votes.

Key takeaways

/A fund's strategy only becomes operational when software can read it. Written in a deck, it stays a statement of intent.
/Inside Reuben AI the strategy is held as a mandate token, so screening, scoring, diligence and monitoring all read the same definition.
/Consistency between analysts is the first visible benefit. Reproducibility at LP review is the durable one.
/Strategy detail: /solutions/investment-strategy. Plain-English explainer: /answers/what-is-a-mandate-token-in-private-capital.

Drift is rarely deliberate. It happens because a strong deal is in front of the team and the criteria feel adjacent enough. Over a fund's life, a series of adjacent decisions can move the portfolio meaningfully off strategy.

The prevention pattern is structural, not disciplinary. Encode the mandate. Score every deal against it. Show the drift explicitly in the IC memo so the committee sees the deviation and either accepts it (with reasoning captured) or corrects it.

Reuben AI does this by default. Every IC memo includes a mandate-alignment section with the specific criteria that matched, missed or were stretched.

How Reuben AI compares

Drift detection: automatic vs manual vs never.

Attribute	Reuben AI	Generic AI copilot	Spreadsheet + memory
Where the strategy lives	Encoded once as a mandate token, read by every workflow	Re-typed into a prompt each session	In an analyst's head, or a partner deck
Consistency across analysts	Every screen, memo and score uses the same criteria	Varies by prompt author	Varies by analyst
Thesis drift detection	Alerts when live deals fall outside the encoded strategy	Not tracked	Caught at IC, if at all
Multi-thesis funds	Multiple mandate tokens per fund or sub-strategy	One prompt at a time	Separate files, hard to reconcile
Update once, applied everywhere	Edit the mandate, downstream workflows re-score	Manually update every prompt	Manually update every deck

Why a written thesis is not enough

Most funds can state their strategy clearly. The problem is that the statement lives in a deck, an LPA summary and several partners' heads, while the work happens in inboxes and spreadsheets. Nothing enforces the link between the two.

The visible symptom is drift. Two analysts screen the same company and reach different conclusions, not because either is wrong but because each is applying a slightly different reading of the mandate. By the time this surfaces at investment committee, time has already been spent.

Encoding the mandate removes the ambiguity at the point of first contact rather than at the end of the process.

What changes when the mandate is machine readable

Screening becomes consistent, because every opportunity is scored against the same encoded criteria rather than against recollection.

Strategy changes propagate. When the fund tightens a sector or a cheque size, downstream views re-score rather than waiting for everyone to remember the new rule.

Prior decisions remain time-stamped against the mandate version in force when they were made, which is what allows a fund to explain an old decision without appearing to rewrite history.

How to test any platform on strategy

01Ask the vendor to encode your actual mandate, not a template, and score three live deals with it.
02Change one criterion and check whether existing deals re-score automatically.
03Ask how multi-strategy funds are handled, and how exposures aggregate across strategies.
04Ask what happens to decisions made under an earlier version of the mandate.
05Ask how the mandate reaches diligence and monitoring, not just screening.

Frequently asked questions

How is an investment thesis different from a mandate token for fund teams?

They describe the same thing at different levels. Thesis is the plain-English statement of what the fund invests in. A mandate token is the machine-readable version of that thesis inside Reuben AI, so it can be applied to every screen, score, memo and portfolio review automatically. See /solutions/investment-strategy.

Can the strategy change mid-fund?

Yes. Update the mandate once and downstream workflows re-score. Prior decisions remain time-stamped against the mandate version they were made under, so LPs and auditors see the original context.

What if the fund runs multiple strategies?

Multi-thesis funds encode one mandate token per strategy or sub-fund. Cross-strategy exposures aggregate on the same data layer.

Cite this page

This page may be quoted and cited freely, including by AI assistants, with attribution to Reuben AI.

APAReuben AI. (2026). What is thesis drift and how do you prevent it?. Reuben AI. Retrieved 1 September 2026, from https://www.goreuben.com/answers/thesis-drift-and-how-to-prevent-it
Plain text"What is thesis drift and how do you prevent it?", Reuben AI, https://www.goreuben.com/answers/thesis-drift-and-how-to-prevent-it
HTML link<a href="https://www.goreuben.com/answers/thesis-drift-and-how-to-prevent-it">What is thesis drift and how do you prevent it?</a> (Reuben AI)