Today only about 10% of decisions at an enterprise are made in a truly data-driven way. There’s massive latent demand to bring the other 90% up to that standard. We’re building the finance data layer to unlock that 10x difference.
For the past year, we’ve been betting on agents getting better at long-running tasks. The two bottlenecks are:
- Building the tooling to extract, load and structure data where judgement needs to be applied
- Collecting and structuring all the business context the agent needs to get the answer right
Traditional software can’t apply judgement, and humans applying judgement doesn’t scale. That’s why we’re building Margin agents: they apply judgement at scale and structure all of an enterprise’s data to unlock 10x more decisions rooted in granular data.
The biggest upside from adopting AI in enterprise finance is in the hundreds of margin-shaping decisions your teams make every day.
How do we price this customer?
Where do we put the next dollar of growth spend?
Which supplier do we choose?
But at almost every enterprise we’ve talked to, the data behind those decisions is messy. Systems don’t reconcile, and costs aren’t tied to the customers and products they belong to. If your best finance analyst can’t get a reliable answer out of it, an AI pointed at your systems won’t either.
margin puts trusted data behind each of your decisions on pricing, growth and ops.
margin’s agents learn how your business works, structure the data across your key systems, and keep it right as the business changes. You get customer- and product-level profit, reliable forecasts of your revenue and costs, and data that reconciles to the ledger.
We’ve spent the last 3 years deploying AI with finance orgs at leading manufacturers and financial institutions in the US and Europe. Two decisions their teams have unlocked with Margin:
- Large segments of users targeted in growth were loss-generating. Redirecting marketing spend improved LTV/CAC by 25%.
- The customer support site that looked the cheapest was 2x more expensive due to low efficiency. Changes to capacity planning were worth $15m in savings within 6 months.
Reach out if you want to deploy AI in enterprise finance but are worried about the quality of your data!
In 3 months sessions will be relics of the past.
- Managing sessions is horrible as a user.
- The bitter lesson for sessions will be that agents perform better if they can orchestrate the sessions and context between them and we are not artificially cutting the latent space with human created session
This is quite obvious if we check where the frontier like @capydotai@cursor_ai projects, @factory etc are going. Directionally apps will look closer to @instinct than to chatgpt.
Eng conversations right now:
- complaining about CI costs
- complaining about Greptile’s pricing being outrageous
- flexing on how good your agent fleet setup is