As a CFO, I've always hated how much of my day is choosing between control and speed. You either get real-time visibility or you get your team moving fast… never both.

My Brex card is different. Brex gives me the controls I need without slowing down important expenses, like paying my podcast producer (shoutout to Ben, our Creative Dictator) or buying Walter treats from Bark Box (which apparently now comes in a bag).

With Brex, receipts are automated, expenses hit in real time, and categorization doesn't require three people and a spreadsheet. I can see exactly where money's going while we focus on growing the Mostly Metrics world empire.

Finance doesn’t have to slow companies down - it can and should push them forward. See why 35,000+ global companies (and me!) use Brex to spend smarter and move faster::

The #1 question I got from people during annual planning season was:

“How the hell do I plan for customer success?”

So I figured screw it, let’s solve it together.

If you take this quick survey, I’ll write about the results at the end of the month. Please take it. It’s really easy (and anonymous, I don’t even ask for your email or anything).

And how we sell that labor calls for a different GTM process.

The biggest challenge people have right now (and it's wicked fun to figure out) is forecasting talent.

If your product is super agentic, it fundamentally changes the roles you need. If your product just works, do you need an SDR, an AE, an SE, a CSM... probably not?

Trying to figure out what we keep vs. toss from the old world of SaaS is hard. Especially because all the roles and ratios are changing.

Do I hire someone who's 40 and has been in SaaS for 20 years and knows the playbook? Or someone who's younger but is a first principles thinker?

In SaaS it was a process. We hired someone who's good at this and someone who's good at that and it was like a bread factory. But now the roles blend.

As someone who was brought up on the playbook of X BDRs and Y SEs per Midmarket AE, the rules of thumb are gone when you're forward deploying engineers.

Which begs the question...

Palantir is the OG when it comes to Forward Deployed Engineers (FDEs, because FTEs weren't cool enough). And TBH the term went completely over my head a couple years back when I first heard it. Rightly or wrongly, I figured it was smoke and mirrors to trade at 80x forward revenue.

Any who… Palantir didn't have a defined ontology data model or a set of tools. So they would go into a target account, understand the requirements, and build a custom layer.

They turned it into a choose your own adventure, ‘yea, of course we can build that. Just let us take a look around first’, kinda deal.

It took a lot of work, but now there was a fox in the hen house and they had a hunting license to build cool shit with a higher chance of the customer buying that cool shit.

And when generative AI came out, whoah, this set up gets supercharged. With that ontology layer you can now generate a lot of dope solutions, much faster.

But now all the AI engineers are using the same term when they talk about selling their products. From Gen AI FP&A products to CRMs.

I was speaking to Brett Queener recently, who founded Bonfire Ventures. He explained that in many of his AI-native companies, they're already deployed as part of the sales process before you close.

In the old world, buyers would imagine what the product could do, evaluate whether it did what you told them it could do, then decide. In an agentic world where you're telling someone your product can do this job? Well, show me it doing the job.

Now, the good news is, if your product is a WTF product - somebody looks at it and goes 'holy shit, wow' - then you don't have four months of dealing with purchasing.

To get deeper on this, I spoke with Varsha Udayabhanu, who runs the finance team at Invisible. They’re a company living by this FDE motion every day in the enterprise AI space.

Here's how she defines what the FDE actually does: they go understand what your specific process looks like, then build a solution using available modules to answer your specific use case.

It sounds simple, but the key insight is that every enterprise problem is an N of 1.

As Varsha put it:

That messy bespoke zone is the sweet spot for FDEs. They're not selling you on a capability, but rather an outcome to a domain specific solution.

Let me give you my take on what an FDE actually is, and how it impacts the way we cost out the P&L.

Think about a classic Systems Engineer (SE) involved in a SaaS sale. They come up with the system requirements based on the customer's environment to say 'this is what good looks like.'

The gist with AI is that it's so flexible, and the use cases are so specific to that customer's domain (or you don't even know how you'll solve their problems yet). But you have these models you can put on top of it and you're confident you can figure it out.

So basically you put engineers, at the cost of your own company, into the sales process and embed them into these companies to build something for them using those models.

If you think about integrations you've historically done for an ERP, you might hire someone like CrossCountry or Accenture to help you implement it. In this new scenario, you're basically doing the implementation (and more) before you even have a signed contract.

Kyle Poyar, my co-host on the Mostly Growth podcast, asked me the $18 question:

Yes. And no. Because they are fundamentally R&D people. But we are doing this before we have a signed contract and we are doing the work to get the customer to say 'Yes, this is an amazing product and I want to keep it.'

This has a number of implications.

On one hand, it sounds really risky to make such an investment in a customer without a dollar commitment. You're putting in a lot of work to get their business. And you're doing stuff that historically companies would charge for.

But here's the thing: Varsha told me that when they walk into enterprises, the business owners have often been sitting with a problem (or tried to have their teams solve it) for 12 to 18 months. They've spent millions of dollars and still haven't seen results.

Invisible's approach lies in solution sprints. They tell the customer:

For example, they'll pick one market and say: let's show you that in this one market, what we're saying works. If you're happy with it, we'll deploy it across every single market. That generally takes six to eight weeks.

What breaks my traditional CFO brain is this: I'm having an outlay of resources ahead of time. Technical resources showing up in my go-to-market costs that wouldn't have been in my CAC payback period before.

But Varsha reframed it for me:

It seems riskier, but the upside is so much higher.

Varsha said something that stuck with me:

The minute you prove you can deliver the outcome they were looking for (that they’ve wrestled with for months or even years), you're in use case one. And immediately, you have a hunting license to find use cases two, three, and four.

She runs the deal desk and thinks a lot about value sold versus value delivered. The reality of enterprise deals is they're multi-millions of dollars. These companies have so many different workflows they've been trying to improve over a long period of time. The opportunity size within most enterprises is massive.

If you can convince them that you'll deliver - that you'll be the partner who helps them sell you to the broader buying committee, who helps them talk about it to five other people in the company who also have problem statements - you're set for multiple years.

Sounds like a good deal for 8 weeks of upfront work.

I asked if Invisible even has SKUs…

Not really.

It's value-based pricing in theory, but she's honest about the reality:

What enterprises actually want is predictability.

They want value, but what they're buying is predictability. As long as you give them a pricing model that's repeatable, where they feel good about paying that much money to you over time… 'I used to spend $10 million on this before, they're going to give it to me for eight, I'm fine'… that works.

Year one, you expect most deals to look like a flat fee. They have a sense for what their current process costs. You replace it with a slightly different cost base. Then as you improve metering over time, you can tie fees to outcomes.

So it’s value-based pricing, but initially anchored as ARR. Then it evolves.

No matter how flexible and extensible AI is (we've all seen the experience of typing into a prompt bar what we want to build and seeing a prototype almost instantaneously), this is a leap of resources. Engineers are expensive AF.

So the cost outlay for the sale you'd think would be higher.

But if this works, the following should prove to be higher:

But here's what's interesting, companies like Invisible are tracking different metrics entirely.

She's focused on what she calls momentum-based metrics. Specifically: use case expansion. Is the customer happy enough with the current use case to give them more? Are they going to push you toward other people in the organization?

And here's a new one: ARR to FDE ratio. What is the productivity of the organization? How many FDEs do you truly need? Is that metric growing or shrinking?

Her framing was cool:

This is why the time-bounded solution sprint matters. You make a disciplined investment, and if it doesn't work, you close the loop quickly.

It begs the question: do you need a classic salesperson to ask for their business anymore?

From Brett Queener again, who, by the way, built the very first quota capacity models at Salesforce:

But for some reason, they didn't want the risk of having a quota.

Because if the product does what it does, it's actually much easier for a traditional seller who wasn't technical to explain and show what the product does. It’s just… there.

In this new world, just to get tactical here: do you think we're still going to give a million dollar quota to someone in mid-market, a $2 million quota to someone in enterprise, and be able to do the same quick back-of-the-envelope math on how many deals they have to get done?

No. In fact it will look more like a ramped deal without a clear commitment amount.

You get into this weird model of: what am I paying the rep for the growth of that firm over time?

Varsha confirmed this is exactly what she's wrestling with. As she put it:

When she thinks about 2026 revenue, it's 'very little of what my AE is going to do.' Because the clients are already found. It's expansion versus new. New is great, but new is really setting up 2027.

The timelines have shifted. Planning is harder. And the atomic unit of headcount planning? They're still figuring it out. But it's increasingly about: how many clients are you in, and what's the expansion potential within each one?

Here's what I keep coming back to: we've moved from buying software to hiring software. And when you hire someone, you don't evaluate them with a demo and a pricing sheet. You see them do the work. You give them a case study.

Varsha gave me one more framing that I think captures where we are:

She's right. The SaaS metrics we grew up on were built for a different model. We haven't found the new back-of-envelope numbers yet. But we're getting closer.

All of this is very new so I don’t have the final box scores on how it will play out. But this is how I’m thinking about it from the perspective of a finance pro who wants to encourage and enable growth:

The bread factory is shutting down soon. We're in the custom bakery era now. And yeah, it's still kinda EZ Bake oven. But it might also be a hell of a lot more defensible once we work the kinks out and figure out how to measure the automated factory lines.

What does it take to raise a Series B these days?

I spoke to a man who wakes up everyday and thinks about the intersection of early venture and growth capital, Jason Kong, General Partner at Base10Ventures.

Base10 invests in “innovation for the real economy”… software built for industries most investors overlook, like many of the blue collar sectors that shape our local communities. If you are passionate about vertical SaaS or back office tools, this is the episode for you.

I’ve written a lot of checks to sales reps over the years (both big and small). Here are eight pillars to apply to your own sales comp plans. They are focused on private companies between $50M and $500M in revenue, but help at any stage.

Hoping you timebound your forward deployment resourcing,

CJ