We surveyed AI adoption at 15 investment banks, the gap between "we have the tools" and "we actually use them" is huge
TLDR: We surveyed L&D leaders at 15 global investment banks. Almost everyone has AI tools, almost nobody is using them effectively for real deal work. Training is too theoretical, nobody knows who's in charge of AI rollout, and 40% of L&D leaders say the pressure to prove ROI is unrealistic. Full breakdown below.
We recently surveyed Learning & Development (L&D) leaders, basically the people responsible for training staff, at 15 global investment banks to understand where AI actually stands heading in 2026. Sharing the highlights here because we think the findings are worth a wider conversation.
Not here to sell anything, just think the data is interesting.
Quick context for anyone newer to finance: "L&D leaders" are the people inside banks who decide what training employees get and how new tools get rolled out. If your firm adopts AI, these are the people figuring out how to actually teach people to use it.
The big finding: having AI tools ≠ actually using them
Almost every bank we spoke to has already purchased AI software, 80% are running Microsoft Copilot (think: AI built into Word, Excel, Outlook). And 80% have increased their AI budget for 2026.
So the money is being spent. The tools are there.
But when we asked how AI is being used for real-deal work, like analysing companies, running models, managing transactions, the picture gets blurrier:
About a third of firms are still just discussing whether and how to use AI
Another third are running small, limited pilots with select teams
Only 3 out of 15 banks are actually scaling AI across multiple deal workflows
So most firms are somewhere between "we're thinking about it" and "we're testing it with a few people." Almost nobody is fully there yet.
The single most consistent complaint we heard: the training people are receiving is too theoretical.
Banks are running general "AI awareness" programs, explaining what AI is, what it can do, why it matters. But that's not translating into people actually changing how they work on live deals.
There's a big difference between understanding that AI can help you analyse a company faster and knowing exactly how to use it when you're in the middle of a real transaction. Most training isn't bridging that gap.
One respondent said it directly: there's almost too much discussion of potential, risks, and governance, and not enough opportunity to just learn how to apply it in practice.
40% of L&D leaders told us the pressure they're under to show measurable productivity gains from AI is unrealistic, given that adoption is still inconsistent across teams and roles.
Leadership wants proof that AI is saving time and improving output. But when half your teams are still in pilot mode, that proof is hard to generate. It's creating a real tension that's slowing things down further.
Based on what we heard, four things stand out:
- Train people for their specific job, not AI in general: a credit analyst and an M&A associate need very different AI skills. Generic training isn't cutting it.
- Make compliance and data privacy practical: employees are hesitant to use AI because they're unsure what's allowed. Clear, simple guardrails would help.
- Pick someone to own it: firms need a clear decision-maker for AI adoption, ideally a joint team across L&D, tech, and compliance.
- Measure what actually matters: time saved, quality of work, fewer revisions, not just whether people logged into the tool.
For those of you working in finance or just starting out, does this match what you're seeing or hearing from firms?