For Investment Bankers
Cognitive Sovereignty Checklist for Investment Bankers
AI tools now generate polished financial models, deal memos, and client presentations so quickly that junior bankers skip the foundational work that used to build their deal instinct. Models that look rigorous can embed assumptions no one on your team understands. Deal memos that are structurally complete can miss the qualitative factors that separate winning deals from expensive mistakes. Your cognitive sovereignty depends on catching these gaps before they reach clients.
These are suggestions. Take what fits, leave the rest.
0 / 19 applicable
Tap once to check, again to mark N/A, again to reset.
Interrogate AI-Built Financial Models Before They Leave Your Desk
Rebuild the revenue bridge without AI assistancebeginner
Take the AI model's revenue forecast and manually walk backwards through customer concentration, churn rates, and contract renewal dates using only source documents. This rebuilds your instinct for what revenue growth actually means in this deal.
Force yourself to state the single biggest assumption in EBITDAintermediate
Write down what operational leverage assumption the model relies on, then find one comparable company that achieved it and one that did not. AI will optimize the middle. You need to know the ranges.
Check whether the model changes materially if working capital swings by 10 percentintermediate
Sponsor-backed companies often understate working capital requirements to make leverage ratios look better. If your model's returns barely move when working capital assumptions shift significantly, the model is hiding risk.
Identify which line items the AI model treats as fixed that should be variablebeginner
AI tends to linearise relationships to simplify the model. Cost of goods sold, distribution costs, and support headcount often scale differently than the model assumes. Walk your deal team through the actual cost structure.
Run a one-page sensitivity on the three assumptions that move valuation the mostbeginner
Do not rely on AI's sensitivity tables. Create your own using only the three variables that matter most for your client's decision. This is what clients remember in the Q&A.
Test the exit assumption by calling three buyers in that sectorintermediate
The model assumes a certain exit multiple or strategic buyer interest. Spend two hours testing whether that assumption holds. You cannot delegate this to AI.
Challenge the capex forecast by asking what actually needs replacingadvanced
AI models capex as a percentage of revenue. In reality, capex is episodic and lumpy. Walk the facility, read the maintenance logs, and build capex from the ground up.
Recover Deal Judgement by Reading the Qualitative Story AI Misses
Read the last three years of customer emails before you read the AI deal memobeginner
AI will see customer concentration in the spreadsheet. It will not see whether customers are threatening to leave, demanding price cuts, or building their own solutions. These conversations determine whether the deal works.
Interview the CFO directly about what keeps them awake at nightintermediate
AI deal memos will miss the executive's unspoken concerns. Spend 30 minutes asking open questions about competitive threats, key person dependencies, and Board pressure. Deal instinct comes from recognising what is not being said.
Identify the one customer or contract the deal depends onbeginner
Every deal has a single customer or contract that makes or breaks returns. Find it yourself before you present. AI will miss this because it lives in narrative emails and verbal agreements, not in spreadsheets.
Map the key person risk explicitly by name, not by functionintermediate
AI will flag key person dependency as a risk factor in the memo. Name the specific person, find out their earnout terms, and check whether they actually stay after close. Your deal could disappear when they leave.
Write down the reason this deal might fail that AI will not mention in due diligenceadvanced
AI generates comprehensive due diligence lists. It will not flag the quiet competitive threat, the regulatory change coming in six months, or the customer mood shift. Spend an hour thinking about what could blow this deal up.
Check whether the sponsor's management team has actually run this type of business beforeintermediate
AI evaluates management track records from CVs and LinkedIn. It cannot tell whether someone truly understands how to run a turnaround or just looks good on paper. Ask them to describe their worst quarter.
Maintain Advisory Differentiation When AI Polishes Your Client Work
Rewrite the executive summary after AI drafts itintermediate
AI will produce something readable and complete. It will not include the insight that actually shifts your client's thinking. Your job is to find one insight that changes how they see the deal.
Add one original comparative analysis that AI cannot generate from public sourcesadvanced
Pull data from your own deal history, your firm's proprietary database, or conversations with other sponsors. This is what clients pay for. Show them something they cannot get from running public comps themselves.
Present the deal case in your own voice instead of the AI-generated tonebeginner
Clients remember how you think about problems, not how polished your slides are. Record yourself talking through the deal for five minutes without notes, then use those words instead of the formal language AI produces.
Identify one assumption in your presentation that differs from what consensus saysadvanced
If your analysis sounds like every other banker's presentation, the client has no reason to choose you. Find one material assumption where you disagree with consensus and defend it in the live presentation.
Design the presentation to show your diligence process, not just the answersintermediate
AI presentations jump to conclusions. Walk your client through the key questions you asked, the places where you disagreed with the company, and how you tested assumptions. This proves you actually did the work.
Prepare to answer the client's pushback on your biggest assumptionbeginner
Do not memorise the AI-drafted language. Prepare three different ways to explain your assumption if the client challenges it. Real advisory means being ready to debate, not just present.
Five things worth remembering
- The most dangerous AI financial models are the ones that look most rigorous. Check whether your junior bankers could defend every assumption if questioned by the client. If they cannot, the model is not finished.
- Your deal instinct lives in the conversations you have now, not in the models you build. Spend the time you save from AI on talking to customers, management, and competitors. That is where you will know what the spreadsheet cannot show.
- When AI generates a deal memo that reads perfectly, it has probably missed the one qualitative factor that determines whether the deal succeeds. Find that factor before you present to your client.
- Rebuilding financial models by hand after AI drafts them feels inefficient. It is not. It is the only way your team builds the judgement to spot mistakes under pressure during a live deal.
- Client loyalty in banking comes from insights they could not generate themselves, not from presentation quality. Make sure at least one page of your pitch shows original thinking that your client did not see coming.
Related reads
Common questions
Should investment bankers rebuild the revenue bridge without ai assistance?
Take the AI model's revenue forecast and manually walk backwards through customer concentration, churn rates, and contract renewal dates using only source documents. This rebuilds your instinct for what revenue growth actually means in this deal.
Should investment bankers force yourself to state the single biggest assumption in ebitda?
Write down what operational leverage assumption the model relies on, then find one comparable company that achieved it and one that did not. AI will optimize the middle. You need to know the ranges.
Should investment bankers check whether the model changes materially if working capital swings by 10 percent?
Sponsor-backed companies often understate working capital requirements to make leverage ratios look better. If your model's returns barely move when working capital assumptions shift significantly, the model is hiding risk.