I’ve had multiple readers this year ask me to do a breakdown on cash management…

Things like… how many bank accounts should I have? And what type?

And guess what?

I’m a (recovering) ppl plzer.

So over the next four weeks we’ll run a series of one click polls to tap into the wisdom of the crowds.

And in a month’s time I’ll deliver to you a full write up on cash management best practices (supercharged by our readers + Brex).

Here’s the first question.

How much of your company's cash is held at its largest banking partner?

Login or Subscribe to participate

Pencils down!

It covers all layers of the finance tech stack, from FP&A to Core Accounting (ERP) to Procurement to AP. And it’s organized by each vendor’s agentic capabilities (Deterministic vs AI Assisted vs Agentic) across the market segments they sell into (SMB, Mid Market, Enterprise).

In addition to the more well known categories (like ERP’s), we also break out new entrants who have burst onto the scene.

And as a new addition to the report this year, we cover AI taking on what were traditionally human roles.

How Do We Measure How Blind We Are?

Sometimes I feel like I'm blind, deaf, and mute, flipping through a sea of chat prompt tabs, waiting for an answer so I can feel productive.

When I asked Peter Pezaris, the CEO of Proxon, “what's happening in finance that doesn't have a name yet”, he said companies "want to go all in on AI. They've got a mandate to deploy it faster, use it more, get better at it, and yet they can't see that. So how do you measure how blind you are? And what do we call that?"

❝

Every CFO has an AI mandate now, and most teams already see the licenses and bills for credits. Many teams are flying blindfolded and feel more productive than ever but still aren't seeing the outcomes they want.

Julio Martínez, CEO & Co-founder, Abacum

I'm running a small business where I can directly measure my output in terms of pieces written, podcasts shipped, and sponsorships booked. The blindness is much more glaring (if that even makes sense) when you're measuring multiple departments.

Without proper record keeping at the event level and observability at the action level, you can’t trace the lineage of financial records back to what caused them.

❝

A debit and credit is always the effect of some cause. It is never the cause itself.

Parker Gilbert, Co-founder & CEO, Numeric

This shift has also made me acutely aware that accuracy and confidence are two different things. You can be accurate but not really confident in how you got there (there are many such cases in using probabilistic black box AI).

It’s kinda like a four box quadrant where nirvana is being Accurate and Confident. The worst box to find yourself in is being Confident but not Accurate.

Accuracy + Confidence = Autonomy

You need both confidence and accuracy for autonomy.

❝

AI doesn't fix bad data or conflicting assumptions. It just helps you reach the wrong conclusions faster.

Eléonore Crespo, Co-founder & CEO, Pigment

We spent centuries building teams, reporting lines, department budgets, performance reviews and bonuses to manage people. None of that exists yet for agents.

Counterintuitively, though, agents might leave you less blind than people do. When a person who owns a process leaves, their reasoning leaves with them. An agent leaves its reasoning behind, available to your company forever.

As someone who's had someone who wasn't all that great at their job, but knew a ton of stuff about how stuff got done (in a mediocre way, but still critical), I do like that.

❝

“Debbie was a C+ accountant but she knew how to bill our top partner correctly.”

In many ways we are just at the seeing phase. See the spend, next control it, then we must optimize it.

❝

By 2031, finance will operate much more like a continuous decision system than a periodic planning function. Finance will also have much richer organizational memory.

Eléonore Crespo, Co-founder & CEO, Pigment

Waiting for Godot

There aren't a ton of benchmarks for what great looks like, and that's assuming we can even get to a distillation of ROI across a budget envelope. To quote Brandon Lewis, head of finance at OpenRouter:

❝

"I know as a [x] billion-dollar business what percentage of our net revenue we should be spending on marketing… That doesn't exist yet on the AI spend side."

It’s admittedly more straightforward when AI is part of what you sell, because you can line it up against product revenue and gross margin. It's much harder for the AI your employees use internally. If we are waiting for Godot in the form of ROI, it will show up for the product-cost bucket long before the employee + OpEx one.

In addition to benchmarks, I also think we are waiting for a wave of churn in 2027. Tools are easy to roll out now, which also makes them easy to rip out. And I don’t think these discounts last forever. There’s a lot of funding floating around the finance tech stack, which is great for finance leaders buying stuff in the short term, but we all understand that in the medium to long term the CAC’s have to CAC and the LTV’s have to, well, live.

To end on a positive note, AI has the potential to expand gross margins and profit margins. But to reap the benefits we need to take a more nuanced view to where we use AI and how we incorporate it with people, and leaders need visibility to verify it’s actually showing up in the P&L.

❝

Agents may execute almost everything. Humans will still decide what is worth doing.

Timothy Wong, VP of Data & AI, Airwallex

There are a lot of market maps in this report. And some will think that being as far to the right as possible is the “best” answer. That’s not necessarily true. The maps show what a product can do, not how much of it you should turn on, and the right spot for a given job depends on how expensive a mistake is.

We want to do better work (humans + technology). Not necessarily AI work in and of itself.

In fact, having AI isn’t good enough. It’s not really a complete assessment.

❝

A lot of people think AI is binary: you either have it or you don't. It's actually a gradient. You can have bad AI, mediocre AI, great AI, or great AI that isn't trained for the use case.

Alex Hagerup, Co-founder & CEO, Vic.ai

It’s like drinking water. Yeah, it’s water, but how clean is it? How sick will this make me?

Stay hydrated, and make sure to call your mom.

Thanks for rocking with me. And I hope this helps you make better buying decisions at your company.

CJ

PS… if you don’t like where your company is, I’d like to formally apologize. I’ve actually hired Conor McGregor to apologize for me:

(jk, email me and I’m sure we can sort it out)

PS PS…

Plz make sure to check out our wonderful partners on this report. My kids need shoes and their feet keep growing.

When your son eats your market map

Reply

Avatar

or to participate