An Investment Analysis Is Incomplete Without the Bear Case
An investment analysis without a bear case isn't analysis. Check AI-generated write-ups against 12 downside categories — dilution, covenants, concentration, and more.
Who this is for
Investment analysts and researchers — Analysts and researchers reviewing AI-generated investment write-ups before they inform a position, a client note, or a committee decision
The problem
An investment analysis that only shows the upside isn't analysis — it's marketing with better formatting. Ask an AI model to summarize a company's prospects and it will typically produce a coherent growth story, because that's the shape of the question most often asked. The specific downside categories that make an analyst's job worth doing — dilution, covenant risk, customer concentration, regulatory exposure — don't reliably appear unless someone asks for them by name.
The cost of this omission is asymmetric. Missing an upside detail costs you a slightly worse entry point. Missing a downside detail costs you the position when the risk actually materializes, at the exact moment you have the least time to react.
How ConvergePanel helps
ConvergePanel runs the same write-up through five models with an explicit downside-category checklist and compares what each model finds, what it omits, and how confident it is in the omission. A category every model addresses with evidence is a genuinely researched risk. A category no model mentions unless prompted is the gap the original analysis left for you to find yourself.
How they compare
| Upside Claim | Downside Factor | Evidence Found | Evidence Omitted | Uncertainty | Potential Impact | Model Disagreement | Reviewer Conclusion |
|---|---|---|---|---|---|---|---|
| Revenue growth funds expansion without new capital | Dilution risk | Cash runway covers 4 quarters at current burn | Convertible notes with a conversion trigger inside the growth timeline | High — depends on share price at trigger date | Material — could increase share count materially | 4 of 5 models never mentioned the convertible structure | Add dilution scenario to the write-up before circulation |
| Growth continues at the current rate | Covenant risk | No mention of credit facility terms in the source document | Leverage covenant tied to EBITDA, not just revenue growth | Moderate — depends on margin trajectory, not just growth | Material if margins compress alongside growth | 3 of 5 models flagged the covenant only after being asked directly | Verify covenant terms and current headroom before relying on the growth case alone |
How it works
- 1Take the AI-generated analysis or write-up as given
- 2Run it against ConvergePanel's downside-category checklist across five models
- 3Compare which categories each model found evidence for versus omitted
- 4Flag categories with no evidence and high uncertainty for direct source review
- 5Add the missing downside case to the write-up before it's used
Use cases
- Auditing an AI-drafted equity write-up before it goes into a client note
- Checking whether a growth narrative addressed dilution and covenant risk
- Confirming a bull case accounted for regulatory or cyclicality exposure
- Building a downside checklist into a standing research review process
Twelve categories worth checking by name
- Liquidity risk — can the plan be funded if a catalyst is delayed
- Refinancing risk — what happens at the next debt maturity
- Customer concentration — how much revenue sits with a small number of accounts
- Regulatory exposure — pending rules or reviews the narrative doesn't mention
- Margin compression — what happens to the thesis if costs rise faster than price
- Dilution — convertible instruments, option pools, or planned raises
- Covenant risk — leverage or coverage triggers tied to the credit facility
- Cyclicality — whether the growth rate assumes a cycle stays favorable
- Execution risk — whether the plan depends on something management hasn't yet done
- Dependency on one catalyst — how much of the return needs one specific event
- Valuation compression — what happens to the multiple if sentiment resets
- Management credibility — whether guidance has a track record of being met
Why the omission is invisible until you name the category
A missing downside category doesn't read as missing — it reads as a complete analysis, because nothing in the text signals what was left out. That's the actual failure mode: not a wrong number, but an absent one, sitting in a blank space the reader has no way to notice without checking for it directly.
Naming the category is what makes the gap visible. Once "dilution" or "covenant risk" is an explicit line item a model has to address, its absence in the original write-up becomes obvious instead of invisible.
Frequently asked questions
Why doesn't an AI model include the downside case by default?
Because the question that produced the write-up usually asked for prospects or an investment case, not a risk audit. The model answers the question it was given — the downside categories have to be requested explicitly to reliably appear.
Is a risk section in the original analysis enough?
Not necessarily. A generic risk section often lists broad categories ("market risk," "competition") without checking whether the specific mechanism — a convertible note, a covenant, a concentrated customer — actually applies to this company. Specificity is what the checklist adds.
What if all five models miss the same downside category?
It happens, particularly for details buried in filings a model may not have deeply processed. That's why the checklist prompts for each category by name rather than relying on the models to surface it unprompted — and why a flagged high-uncertainty category still needs a direct source check.
Does finding an omission mean the upside case is wrong?
No. An omitted downside category is a gap in the analysis, not proof the conclusion is incorrect. It means the case hasn't been tested against that specific risk yet — the next step is checking it, not discarding the thesis.
Can ConvergePanel calculate the financial impact of a missing risk?
No. It identifies which downside categories the analysis addressed versus omitted and flags the uncertainty around each. Quantifying impact — sizing a covenant breach or a dilution scenario — is financial analysis that requires a qualified professional.
Explore related pages
ConvergePanel provides AI-assisted verification for informational purposes only. Not forensic analysis. Not legal evidence.
More in Research
Deep Research with Multiple AI Models
Run complex research questions through 5 AI models at once. ConvergePanel synthesizes consensus, disagreements, and bias signals into one structured brief.
Compare ChatGPT, Claude, Gemini, Grok, and Perplexity for Research
Compare ChatGPT, Claude, Gemini, Grok, and Perplexity for research. Learn when models agree, disagree, miss context, or need verification.
AI Research for Decision-Making Teams
Decision-making teams need shared, reliable research inputs. Multi-model AI surfaces consensus, disagreements, and uncertainty — not just one AI's take.