What should an investment committee framework include?

It should include the mandate and eligible universe, the scorecard version, primary evidence, the thesis and disconfirming case, position sizing and portfolio fit, conflicts and exceptions, approval conditions, monitoring triggers, owners and the next review date. The decision should remain reproducible after the data or model changes.

How can an investment committee be rule-based without becoming rigid?

Apply rules to the process rather than forcing every conclusion. Define what evidence must be reviewed, how risks and position limits are checked, who may override a model, what reason accompanies an override and when the decision is revisited. The committee retains judgement but cannot change the process invisibly.

What is an investment committee decision record?

It is a dated record of the information and process behind an approval, rejection or deferral. It should preserve the security, mandate, data vintage, scorecard version, thesis, key risks, evidence links, approved size, exceptions, monitoring conditions, participants and review date.

Should an investment committee use AI?

AI can retrieve passages, compare calls, organise the evidence pack and draft a summary. It should not invent missing facts or silently calculate ratios, scores or forecasts. Material claims should cite the source, quantitative work should be deterministic, and the committee should remain responsible for the decision.

A Rule-Based Investment Committee Framework for Indian PMS, AIF and Family Offices

A rule-based investment committee does not replace debate with a score. It gives every decision the same minimum structure: mandate, model version, primary evidence, bear case, portfolio fit, approved size, exceptions, monitoring triggers and a dated record. The committee keeps its judgement, but it cannot quietly change the rules or rewrite the evidence after the outcome is known.

For an Indian PMS, AIF or family office, that is the practical meaning of portfolio governance.

Why committee process fails even when the people are good

Most committee failures are not caused by a lack of intelligence. They come from inconsistent process:

one analyst brings a full evidence pack while another brings a persuasive story;
a score is quoted without the formula or the data date;
the company is discussed in isolation from the existing portfolio;
an exception is agreed verbally and never recorded;
approval conditions do not become monitoring rules;
the original thesis is replaced by a new explanation after the stock moves;
a revised scorecard changes what an old decision appears to have seen.

The purpose of a framework is not to make every meeting identical. It is to keep the load-bearing questions from disappearing when conviction is high or time is short.

The rule: govern the decision path, not the conclusion

A committee should be free to approve a low-ranked company or reject a high-ranked one. A quantitative model cannot see every regulatory, accounting, management or competitive fact.

The governance requirement is that the exception becomes visible:

Which rule or score was overridden?
Who made the judgement?
What evidence supported it?
What condition would make the exception invalid?
When will the committee review it again?

This is flexible where judgement adds value and rigid where memory creates risk.

The eight-part committee pack

1. Mandate fit

Begin with eligibility before discussing attractiveness.

The pack should state the relevant strategy or client mandate, eligible universe, liquidity rule, security type, sector or exposure limits and any prohibited conditions. If the company requires an exception merely to enter the universe, surface that before the pitch begins.

2. Data and model identity

Record the data cut-off date and the exact version of every scorecard or model used. Financial values should reflect what was knowable on that date, not a later restatement.

The committee should be able to answer:

Was the result consolidated or standalone?
Which reporting period and filing vintage were used?
Were corporate actions handled?
Which peer group produced the percentile?
What happened when a factor was unavailable?

If the answer is “the current dashboard,” the decision is not reproducible.

3. Quantitative case

The quantitative section should be compact. Show the factors that matter to the strategy, the company’s rank against the correct peers, the valuation and risk measures, and which hard gates passed or failed.

Do not turn the pack into a data dump. A committee needs the driver of the score and the sensitivity of the result. If changing one threshold or excluding one outlier moves the company from the top decile to the middle, that fragility is decision-relevant.

4. Qualitative evidence

The qualitative section should explain the business, not decorate the score.

At minimum, cover:

the revenue and profit engine;
industry structure and competitive mechanism;
management’s relevant commitments and delivery history;
capital allocation;
accounting and governance questions;
the strongest evidence against the thesis.

Each factual claim should link to a filing, transcript, presentation or other identifiable source. An AI summary without the source is not committee evidence.

5. Variant view and valuation

State what the team believes that is different from the expectation already in the price. Then translate that view into a small set of assumptions rather than one target number.

The committee should see which assumptions carry the valuation, what range is plausible and what happens if the variant view is wrong. A precise price target built on hidden assumptions is less useful than a transparent sensitivity table.

6. Portfolio fit and sizing

Approval and position size should be separate votes or explicit decisions.

Review:

current exposure to the same sector, factor and macro drivers;
overlap through funds or related companies;
liquidity and capacity at the proposed weight;
downside under the bear case;
initial, maximum and review weights;
the benchmark and risk budget affected.

A good company can still be the wrong addition to a portfolio that already owns the same risk several times.

7. Decision and exception record

Use a clear outcome: approve, approve with conditions, defer, reject or exit. Record participants, conflicts, dissent and every exception to the normal process.

Avoid minutes that merely say “discussion held.” A useful record says what was decided and why.

8. Monitoring contract

Every approval should end with three to seven watchpoints. Each needs a condition, evidence source, owner and response.

Examples:

if working capital exceeds the agreed range, the analyst re-underwrites cash conversion;
if management changes annual guidance, the model and committee note are updated;
if promoter pledge rises above the firm’s limit, the position returns to committee;
if factor or sector concentration breaches the portfolio cap, sizing is reviewed;
if a thesis driver cannot be measured on time, the gap is recorded rather than assumed unchanged.

The monitoring contract is what turns a committee from a one-day event into portfolio governance.

A one-page decision record

The following template is deliberately compact.

Field	What to preserve
Decision ID	Stable identifier and date
Security and mandate	Company, strategy and eligible-universe rule
Data vintage	Latest information the committee could have known
Model identity	Scorecard, factor and valuation-model versions
Quantitative summary	Gates, rank, important factors, valuation and risk
Thesis	The mechanism and variant view in two or three sentences
Disconfirming case	Strongest evidence against the thesis
Source pack	Filing, transcript, presentation and calculation links
Portfolio fit	Existing related exposure and concentration impact
Decision	Approve, conditional, defer, reject or exit
Size	Initial weight, maximum weight and sizing rationale
Exceptions	Rule overridden, author, reason and evidence
Watchpoints	Condition, source, owner and required response
Review	Next date or triggering event

The record should be exportable and human-readable. A firm should be able to open it without the original analyst in the room.

What should stay deterministic

The following should be computed by explicit rules:

financial ratios and growth rates;
factor percentiles and composite scores;
portfolio weights and concentration;
backtest returns and risk statistics;
threshold tests and alert conditions.

The following can involve judgement, but should remain sourced and dated:

moat and industry interpretation;
management credibility;
the variant view;
the relevance of a one-off event;
the decision to override a score;
conviction and the final allocation within approved limits.

This separation prevents a language model from producing a persuasive number and prevents a formula from pretending it understands context.

Where AI belongs in the committee workflow

AI can improve preparation materially:

retrieve every passage relevant to one assumption;
compare current and prior calls;
list changes in guidance;
run the same diligence question across several companies;
draft the evidence index and meeting summary;
connect a new filing to existing thesis watchpoints.

It should be fenced by the source pack. If evidence is absent, the output should say so. Ratios, scores and forecasts should remain outside the language model and be reproducible in code or Excel.

The scarce part of a committee is not memo writing. It is deciding which assumptions deserve belief and how much capital that belief deserves.

How Altys supports this framework

Altys is built to keep the committee inputs and the later monitoring record on the same evidence trail. A team can define its own screens and factor scorecards, replay sealed rules on point-in-time data, investigate the shortlist through source-linked company evidence, preserve the decision, and turn thesis conditions into monitoring rules.

Important calculations and outputs can be exported to Excel with formulas so the committee can validate the work outside the platform. Altys does not ask the committee to accept a hidden score and does not place the trade. The firm owns the rule and the judgement.

A monthly governance review

The committee should review the process as well as the positions. Once a month, ask:

Which alerts fired, and which were useful?
Which thesis assumptions changed without triggering a review?
How many model overrides occurred?
Were overrides concentrated in one analyst, sector or factor?
Which required inputs were missing?
Did any live portfolio breach differ from the written mandate?
Did a new scorecard version materially change old classifications?
Which forecast or management promise was wrong, and what should the process learn?

This is where a rule-based framework compounds. It turns exceptions and mistakes into observable process data.

The principle

An investment committee is not rigorous because it meets regularly or produces a long memo. It is rigorous when another person can reconstruct the decision, challenge the evidence and see what the portfolio agreed to watch.

Rules provide the spine. Evidence supplies the facts. Judgement still makes the decision.

This article is educational and describes a possible operating framework, not legal, compliance or investment advice. A firm’s actual obligations depend on its registration, mandate and advisers. Altys Labs is not a SEBI-registered Research Analyst or Investment Adviser.