Investment bankers have been promised automation for years.
This time, a meaningful amount of the work really is being automated.
AI can already help with company research, comparable-company analysis, document review, diligence, presentation drafting, market mapping and first-pass financial work.
In a widely cited 2023 analysis, Deloitte projected that generative AI could improve productivity in investment-banking divisions by an average of 34% by 2026, with M&A, issuance and advisory among the areas likely to benefit. That was a forecast rather than a measured 2026 outcome, but the direction has proved hard to dismiss.
Look at Rogo. The company was started by former bankers who wanted to automate some of the work they had been doing themselves. In April it raised $160 million in a Series D at a reported $2 billion valuation. Rogo says it is deployed across leading investment banks, asset managers and private-equity firms. Kleiner Perkins says more than 35,000 bankers and investors use the platform across firms including Rothschild, Jefferies, Lazard, Moelis and Nomura.
AI in investment banking is no longer a theoretical debate.
The more interesting question is what happens when everyone has access to much better tools.
The first draft is becoming much cheaper
Investment banking has traditionally required huge amounts of human effort to collect, organize, check and present information. That is exactly the kind of work AI can compress.
Research that once took hours can take minutes. An initial buyer list can be produced almost instantly. A first version of a pitchbook, market map or industry analysis can be generated far faster.
None of this means the output can simply be sent to a client. Investment banks still need controlled systems, appropriate handling of confidential information, auditability and human verification. Hallucinations, bad source data or an incorrect valuation are very different problems in a live transaction from an error in a casual AI query. Deloitte itself has highlighted legal, reputational and operational risks alongside the productivity opportunity.
But the economics of producing the first draft are changing. That has consequences for both the junior-banker model and what clients should be willing to pay for.
A transaction is not primarily an information problem
The difficult part of an important M&A process often begins after everyone has the spreadsheet.
A founder wants to sell but is not emotionally ready. The board disagrees on timing. The obvious strategic buyer will not engage. A bidder likes the company but cannot get the valuation through its investment committee. Diligence produces an unexpected problem. The leading buyer tries to retrade. Management loses confidence in the process.
These are not primarily information problems. They are judgment and relationship problems.
The senior banker needs to understand the motivations of the people involved, decide when to push and when not to, preserve competitive tension, get someone back to the table and know when a process has genuinely reached its limit.
AI can prepare a banker extremely well for those conversations. The banker still has to have them.
Silicon Valley technology bankers face a second AI challenge
Bankers are not only adopting AI inside their own firms. They are advising companies whose economics are being changed by it.
Reuters reported in July that major Wall Street banks see an AI-driven capital-expenditure “super cycle” creating substantial financing and deal activity across equity, debt and M&A.
At the same time, AI is making some technology businesses harder to underwrite. Reuters Breakingviews recently highlighted growing caution among lenders toward parts of the software market as investors try to distinguish businesses that AI could strengthen from those it could weaken.
That makes “technology investment banking” a broader and more technically demanding category. An AI-infrastructure banker may need to understand data centers, chips, networking, power and extraordinary capital requirements. A software banker needs a view on whether AI strengthens a company's product, compresses pricing, lowers barriers to entry or changes the value of its proprietary data. An AI-native company may have economics that look very different from the SaaS businesses investors became accustomed to valuing over the previous decade.
Deloitte now argues that AI itself needs to become part of M&A diligence because it can materially affect competitive position, revenue, costs, margins and valuation. The banker has to understand that well enough to challenge an AI-generated answer rather than simply accept one.
Relationships may become more valuable, not less
If every banker can produce competent research more quickly, research itself becomes less differentiating. Everyone can identify the same company. Everyone can generate a long list of potential buyers. Everyone can summarize an earnings call.
What remains much harder to reproduce is knowing why a buyer might act, who inside the organization matters, what happened the last time they considered a similar transaction and whether a particular CEO is likely to take a call.
The same applies to private equity. Knowing that a fund invests in software is public information. Knowing what its partners are genuinely trying to buy this quarter is not.
AI could therefore remove some of the production advantage historically associated with very large junior teams. That does not automatically mean boutiques win. Large banks have proprietary data, enormous technology budgets and internal systems that smaller firms cannot easily replicate. Deloitte has explicitly noted both possibilities: AI could reduce some barriers to entry while the investment required to build sophisticated systems could also widen the gap between large institutions and smaller firms.
It is too early to know which effect will dominate. What seems clearer is that mediocre advice becomes harder to justify when basic analytical work gets cheaper.
Junior banking should change
I do not think this necessarily means eliminating junior bankers. There is a better outcome.
Less time rebuilding information that already exists. More time checking whether it is correct. Less time formatting. More time understanding why the analysis matters. Earlier exposure to clients. Earlier exposure to the decisions that actually determine whether a deal works.
That only happens if banks continue to teach judgment instead of allowing software to become a substitute for learning it.
What remains worth paying for
At Open Future Forum, we spend a lot of time with CEOs, CFOs, investors, private-equity professionals and technology leaders dealing with AI from different perspectives. Investment bankers sit in an especially interesting position because they are close to the moments when companies are financed, bought and sold.
For senior technology bankers in Silicon Valley, I don't think the question is whether to use AI. That decision has largely been made. The harder question is: which parts of investment banking should become dramatically faster, and which parts become more valuable precisely because they cannot be automated easily?
I expect a great deal of production work to change. I do not expect a founder who has spent 15 years building a company to outsource the most important negotiation of their career to an agent.
AI can draft the pitchbook. It can help build the model. It can find the buyers.
When the transaction becomes difficult, somebody still has to run the deal.
Last updated: August 23, 2026
Frequently Asked Questions
Where dealmakers work through AI's real impact
Open Future Forum convenes CEOs, CFOs, investors and technology leaders on how AI is actually changing how companies are financed, bought and sold. Global, founded in Silicon Valley.