Drafting Equity Research Reports with AI

The sell-side equity research report has a fixed structure: thesis, valuation, financials, risks, catalysts, ESG considerations, and disclosures. Sell-side analysts know that the harder part is not writing the report — it is having a defensible view. Buy-side analysts produce a leaner version of the same structure for internal memos. In both cases, AI can produce a strong first draft in 30 minutes that you spend the next 90 minutes sharpening.

This lesson covers the full draft-to-publish workflow.

What You'll Learn

• How to translate your model and notes into a structured research report
• How to write a defensible bull/bear/base case
• How to draft compliance-friendly language
• How to produce different report formats (one-pager, full initiation, post-print update)

The Layered Drafting Approach

The single best technique: do not ask AI for a full report in one prompt. Layer the work.

Layer 1: Outline. Get the structure right before generating prose.

I am writing a 10-page initiation report on [COMPANY] for a sell-side
research team covering [SECTOR]. Generate the outline. Use standard
sell-side structure with these required sections:

- One-page executive summary
- Investment thesis (3-4 bullets)
- Valuation summary
- Company overview and history
- Industry overview
- Financial summary (5 years historical, 3 years forecast)
- Bull/Bear/Base case scenarios
- Catalysts
- Risks
- ESG considerations
- Management
- Disclosures

For each section, specify length (in words or paragraphs) and what
must be included.

Layer 2: Section drafts. Once the outline is approved by your senior, draft each section in its own prompt with full context.

Layer 3: Tightening. After all sections are drafted, paste them back into AI for global edits — consistency of tone, terminology, and the through-line of the thesis.

Writing the Investment Thesis

The thesis is the spine of the report. If it is weak, nothing else matters. Use this prompt:

Help me sharpen the investment thesis for [COMPANY], rating [BUY/HOLD/SELL]
with a price target of $[X].

My current thesis bullets are:
1. [BULLET 1]
2. [BULLET 2]
3. [BULLET 3]

For each bullet, do the following:

- Rewrite to lead with the variant view (what we believe that
  consensus does not)
- State the supporting evidence in one sentence
- State the timeframe over which we expect it to play out
- Flag what could invalidate the thesis

Each rewritten bullet should be 35-45 words. No filler. No "we believe"
or "in our view" — write declaratively.

The "variant view" framing matters. A thesis that matches consensus is not a thesis — it is a description. Forcing the AI to find the variant view sharpens your own thinking.

Drafting Bull, Base, and Bear Cases

For [COMPANY], draft the bull, base, and bear case writeups for
our [BUY/HOLD/SELL] rating with a base case PT of $[X].

For each case:

- Specify the implied 12-month price target
- State the three to four assumptions that differentiate this case
  from the other two (use these specific drivers: revenue growth,
  margins, valuation multiple, capital allocation)
- Quantify the probability we assign to this case
- Explain the catalyst path that would lead to this case

Format as three sections of equal length (about 120 words each).
Sell-side research tone. Make sure the bull case sounds achievable
and the bear case sounds plausible — neither should sound like a
straw man.

The "neither should sound like a straw man" instruction is critical. Junior analysts often write bear cases that no one would believe — that destroys the credibility of the bull case too.

Industry Overview Sections

Industry overviews are tedious and AI handles them well:

Draft the industry overview section of my [COMPANY] initiation report.
The company operates in [SECTOR/SUB-SECTOR]. Cover:

1. Market size and growth (cite a recent industry report — Gartner,
   IDC, McKinsey, or sector trade body — and the methodology)
2. Key value chain players (suppliers, competitors, customers)
3. Competitive dynamics (consolidating? fragmented? regulated?)
4. Three structural trends that will shape the industry over five years
5. Cyclicality and macro sensitivity

500 words. Cite sources by name. Where you are unsure, write "[need
to verify]" instead of inventing a statistic.

The "[need to verify]" trick is a guard against hallucination. AI will mark its uncertain claims and you can fix them with a quick Perplexity search.

Writing About Risks

Risk sections are where compliance gets nervous about AI. Tight prompting helps:

Draft the risk factors section for [COMPANY]. Group risks into:

1. Macro risks (specific to this business, not generic)
2. Company-specific risks (concentration, key man, technology, etc.)
3. Regulatory and legal risks
4. ESG risks
5. Execution risks against management's stated strategy

For each risk:
- One-sentence description
- Quantified impact if possible (e.g. "10% decline in customer X's
  spend = $40M revenue impact based on FY24 disclosure")
- How we have already factored it into the forecast

Avoid boilerplate. If a risk would apply to any company in the
industry, only include it if it is unusually acute for this company.

The "avoid boilerplate" rule is what differentiates a useful risk section from a legal disclaimer.

Catalysts and Triggers

Draft the catalysts section. Group into:

- NEAR-TERM (next 3 months)
- MEDIUM-TERM (next 6-12 months)
- LONG-TERM (12+ months)

For each catalyst, specify:
- Event description
- Expected date or window
- Direction of impact (positive / negative / neutral)
- Magnitude (low / medium / high)
- What we will be watching to confirm the thesis

Investors care about timing. Be specific. If you do not know the
date, say "TBA — typically announced in Q[X]" rather than guessing.

Compliance-Friendly Language

Most sell-side firms have specific language requirements: no use of words like "guaranteed," disclosure of conflicts, balance of positives and negatives. AI is good at applying these:

The attached draft section is currently bullish in tone. Rewrite
to comply with sell-side research norms:

- No words: guaranteed, will, certain, definitely, no risk
- For every positive claim, include the relevant uncertainty
- Use "we expect," "we estimate," "we believe" appropriately
- Include the standard valuation methodology language
- Flag any forward-looking statement so I can apply standard
  forward-looking disclosure

Keep the substantive view unchanged.

This is not optional — your firm's compliance team has rejected research over the words "guaranteed" and "will" in the past. AI catches them faster than you do.

The One-Pager Format

A different beast from the full report. Compress aggressively:

Reduce my full initiation report on [COMPANY] to a one-page summary
suitable for portfolio manager review. Use this structure:

- TICKER, rating, price target, expected return %
- The one-sentence thesis
- Three bullets of supporting evidence
- One bullet of key risk
- Valuation summary (one number per method)
- Two near-term catalysts

Total: 250 words maximum. No section headers. Dense and scannable.

PMs read one-pagers; they rarely read the full report. The one-pager is the marketing. When that one-pager has to become an actual pitch deck or board exhibit — for an internal IC presentation or an investor-relations-style readout — the slide-craft (headline-test titles, building the exhibits, narrative flow) is its own discipline, covered in AI-Powered Pitch Decks & Executive Presentations.

Buy-Side Memo Format

Buy-side memos are leaner and more opinionated than sell-side reports. The prompt changes:

Draft the investment memo for [COMPANY] for our PM committee.
We are pitching a [LONG/SHORT] with a [SIZE]% position. Sections:

- The pitch (the variant view in two sentences)
- Why now (what changed in the last 90 days)
- Path to revaluation (catalysts in next 12 months)
- What we own that consensus does not
- Bear case (what could go wrong and how much we lose)
- Position sizing and exit plan

Total: 600 words. Direct, opinionated tone. No sell-side hedging.

Key Takeaways

• Draft in layers: outline first, then sections, then global edit
• Sharpen the thesis by forcing variant views and timeframes
• Bull/Base/Bear cases all need to be plausible — no straw men
• Mark uncertain claims with "[need to verify]" so you fix them
• Compliance language is a checklist AI handles reliably — but you still own the final document