AI Enabled Investment Banks

The future of services is Language Models

.

Two years ago, whilst working as an intern at a VC fund - I was asked to map the AI landscape.

My conclusions were:

1. The foundation model layer was too capital intensive to invest

2. NVIDIA had a stranglehold on the chip layer

3. AI applications going after domains previously dominated by text / paperwork (Accounting, Law, Finance)

The company I thought was most interesting at the time … Harvey AI!

At the time they had just raised a Series A from Sequoia. Today, Harvey is worth >$3bn and is onto their series D. I should have tried to invest somehow…

Regardless of this miss, I think #3 is even more true today.

So where is the opportunity in AI services today?

Services > Software

I think building a services company is potentially a more interesting play than building a software company itself.

There are lots of AI-software companies selling into existing professional services companies (Harvey, Hebbia, Rogo, Glean).

But increasingly, this layer feels extremely competitive - with many companies going after the same buyer set. The challenge will be that firms will have an ‘AI strategy’ - and will likely only have budget to purchase from ~1 or 2 vendors.

This means value will concentrate amongst the existing market leaders. The train has likely already left the station here…

I think there is a potentially bigger opportunity to build a services company.

Why?

Existing professional services firms are never going to upend their entire business model and replace it with AI.

Do you really think IB’s / Law Firms will stop hiring analysts and start using models?

I don’t think so.

This is because models are 90% good enough today - but clients need 98% good enough - so firms won’t adopt the tech.

The reality is though in 3-5 years the models will get to 99%+ and be far better than analysts - particularly if you can train them on domain specific data.

If you want to rethink the entire model - you need to start fresh.

This size of the prize is huge. GS made ~$83bn in revenue at a ~15% margin over the last twelve months. The vast majority of that cost is labour. But what if you could have half of the employees?

Could you get to 40%-50% and start to look more like tech companies?

I think so.

Enter Offdeal

I stumbled across a business out of the 2024 Y-Combinator batch called ‘Offdeal’.

Their value prop?

An AI-enabled investment bank.

Rethinking all of the workflows inside an existing bank from the ground up, combined with the white glove service and M&A acumen of a boutique IB.

Their results are already impressive - running 8-10 sellside processes in deal teams of 1-2 bankers, removing analysts / associates entirely.

But why build from the ground up vs try to implement in existing businesses?

There are lots of interesting companies trying to roll-up businesses with AI. Legendary VC Elad Gil just announced he is investing in AI rollups, Thrive and GC are also in the game with Crete and Atticus respectively - AI accounting plays.

But I think starting from the ground up is far more interesting.

The reason is because many of the workflows that you would traditionally perform in these industries is ‘busy-work’ created by misaligned incentives.

After all - you need to justify doing ‘enough work’ for the fees to make sense - so you keep your analysts busy even if the work isn’t high value.

These incentives will change with AI if you don’t have an army of analysts at your disposal.

Taking buyer screening for instance - a job where traditionally analysts would spend weeks trawling the universe for potential bidders, Offdeal can automate via web scraping + email outreach with LLMs to contact bidders and can get LOI’s the same day.

This is the future.

Offdeal has the right GTM strategy going after the SMB market first - the stakes are lower serving SMB customers, and customers will be more willing to take a risk with a new firm.

Where is the Opportunity?

I think you could take the Offdeal playbook and replicate it across law, accounting, or any other white glove professional services vertical.

Whilst the software layer will likely have a small number of big winners, direct services providers are likely to be more fragmented.

If anyone is building in this space - I’d love to have a chat / learn more!

---

That’s a wrap everyone!

If you liked the article, drop a comment below and share it with friends - it means a lot and gives me motivations to keep going :)

Until next time,

Wrucky