For Investment Bankers

20 Practical Ideas for Investment Bankers to Stay Cognitively Sovereign

AI financial models look rigorous but hide assumptions junior bankers cannot interrogate or challenge. When you stop building models yourself, your instinct for what makes deals actually work disappears.

These are suggestions. Take what fits, leave the rest.

Download printable PDF

Building and Testing Models

Hand-build the DCF before checking AI outputbeginner

Build your own valuation first. Then compare against ChatGPT or Kensho output to spot buried assumptions.

Force yourself to state every assumption aloudbeginner

Before running any Bloomberg AI model, write down your assumptions on paper. Read them back. Listen for what feels wrong.

Change one input and predict the output changeintermediate

Adjust a single variable in your model. Estimate the impact before running it. Check if your intuition matches reality.

Challenge the comps AI selected for youintermediate

When Capital IQ AI picks comparable companies, ask why each one belongs. Reject at least two before you accept the set.

Build sensitivity tables without AI assistanceintermediate

Create your own two-way tables by hand. Understand where value lives. Then use AI to speed up the tedious work.

Audit the drivers behind working capital assumptionsintermediate

AI often embeds industry-standard working capital percentages that miss this deal's specific cash conversion cycle reality.

Spot-check three random cells in AI modelsbeginner

Pick cells without telling anyone. Recalculate them manually. Catch where AI has made basic arithmetic or logic errors.

Recreate the leverage bridge without calculatorbeginner

Walk through how debt moves quarter to quarter. Without AI. This teaches you what covenant pressure actually feels like.

Write down what would break this valuationintermediate

Before presenting to client, list three factors that would cut your valuation in half. These are your actual risks.

Ask which assumptions changed from last yearundefined

Compare your new model to last year's. Ask ChatGPT to list changes. Question whether each one reflects real market shift.

Judgment in Deal Work

Write your deal memo before AI drafts itintermediate

First draft your own assessment of why this deal works or fails. Then read what Copilot produced. Notice what you missed.

Identify the single deal killer yourself firstintermediate

Before running due diligence AI, decide which one thing could kill the deal. Use AI to test that, not to find it.

Interview the business owner before analysing databeginner

Talk to them. Take notes. Form your own view of management quality. Then cross-check against what Kensho finds in the numbers.

Document your deal instinct before data reviewbeginner

Write down your gut reaction to the opportunity in one paragraph. Lock it away. Review it after AI analysis is done.

Push back on AI sources during due diligenceintermediate

When AI flags a risk from third-party data, ask where it came from. Verify it yourself. Don't inherit other analysts' mistakes.

Spot the qualitative factor AI cannot quantifyintermediate

Identify one aspect of deal quality that no model captures: founder commitment, supply chain resilience, customer concentration fear.

Compare your price view to AI valuation firstintermediate

Set your own price range based on sector feel and recent deals. Then run AI models. Use the gap to understand your blind spots.

Rewrite the investment thesis AI generatesbeginner

Let Copilot draft your thesis. Rewrite it entirely in your own words. Your version should reflect what you actually believe.

Track your past predictions against AI outcomesbeginner

Keep a record of your deal calls versus what AI recommended. Over time, you will see where your instinct adds value.

Tell the client why you disagreed with AIundefined

If your analysis diverges from AI output, explain the difference to the client. This becomes the part of your advisory that cannot be automated.

Five things worth remembering

The junior banker who still builds models by hand will have better instinct in five years.
If your presentation is 90% AI-drafted, clients have no reason to pay for you next year.
Disagreement with AI output is a signal to investigate deeper, not to trust the machine.
The financial model that teaches you something is worth more than the one that saves you time.
Your deal instinct comes from failing on deals you thought would work. AI lets you avoid that learning.

20 Practical Ideas for Investment Bankers to Stay Cognitively Sovereign

Building and Testing Models

Judgment in Deal Work

Cognitive Sovereignty: How To Think For Yourself When AI Thinks For You