The Stat That Should Change How You Think About Your Team

$1,000 a Day, Three People, No Code Written by Hand

A software company called StrongDM made news for a straightforward reason: their three-person engineering team spends $1,000 a day on AI compute and produces what required a ten-person team eighteen months ago. No human on the team writes code. No human reviews code. The agents build it, test it, and ship it.

That's a number worth sitting with: three people, eighteen months, a tenfold reduction in headcount for the same output.

This story is easy to dismiss as a Silicon Valley edge case. Software is different. Their situation is unusual. Your business isn't a tech company.

But the shift that made this possible isn't specific to software development. Intelligence — the kind that processes information, follows patterns, and makes routine decisions — has become a purchasable commodity. And that changes the math on every business that pays people to do work that follows a pattern.

What Actually Changed

For sixty years, the fundamental unit of work was human time. You hired people, gave them tools, and the constraint on your output was hours of skilled labor. To do more, you hired more.

That's not the world anymore. The cost of what you might call a unit of intelligence — the ability to read information, make a judgment, take an action — has dropped from $20 per million tokens in late 2022 to about $3 today. It's falling another 10 to 200 times annually by most credible projections. You're not buying hours. You're buying intelligence at a price that keeps dropping.

Here's the counterintuitive part: when a resource gets dramatically cheaper, you don't use less of it. You use orders of magnitude more. This has been true of every resource in computing history. Cloud storage got cheaper and companies stored vastly more data, not less. Bandwidth got cheaper and video streaming exploded. Processors got more efficient and software got more complex, not simpler.

Intelligence is following the same curve. The average company now spends $85,000 a month on AI, up 36% year-over-year. Forty-five percent are planning to spend over $100,000 monthly. This isn't companies replacing a few subscriptions with a better tool. This is organizations restructuring what they spend on intelligence.

For StrongDM, $1,000 a day in compute for three engineers is still cheaper than the ten-person team it replaced. That's why the math works. And the compute keeps getting cheaper while the output keeps getting better.

What Other Companies Are Already Doing

StrongDM is an extreme example, but the pattern shows up at every scale.

Shopify announced in early 2025 that they would evaluate headcount additions by asking whether an AI agent could do the job first. That's not a cost-cutting posture — it's a default operating assumption that intelligence is now available outside of human labor. Their merchant support operation now routes the majority of incoming queries through agents before any human involvement. The humans handle exceptions, escalations, and the situations the agents flag as outside their scope.

HubSpot has publicly discussed using AI agents to handle initial lead qualification, routing only the leads that meet defined criteria to their sales team. The agents don't close deals. They do the sorting work that used to take a BDR hours every week — reviewing hundreds of inbound leads to find the thirty worth calling. That's a pure effort problem, and agents do it faster and more consistently than people.

The pattern is the same in both cases. The intelligence handles the pattern work. The people handle what the pattern can't cover. And the ratio of pattern work to judgment work in most service businesses is much higher than owners typically realize until they actually map it.

What This Actually Means for a 10-Person Service Business

You're not trying to get to $3 million in revenue per employee. But here's what does apply to a roofing company, a logistics operation, a mortgage brokerage, or an HVAC service firm with ten people on staff.

Your team has two categories of work. They may not think about it this way, but the categories are real.

The first category is what you'd call effort work — high-volume, repetitive, pattern-based tasks that follow the same logic every time. Intake forms. Follow-up sequences. Scheduling confirmations. Invoice processing. Status updates. Data entry. This work isn't intellectually demanding. It's demanding because there's a lot of it, it has to happen consistently, and it takes real hours away from everything else.

The second category is judgment work — understanding what a client actually needs, reading a situation that doesn't fit the standard script, deciding whether to bend a policy, knowing when a long-term customer relationship is at risk. This work requires the context and experience your team has built over years. It can't be reduced to a pattern.

AI handles effort work. Not eventually — now, at a price that makes sense for a business your size. The cost to run a follow-up agent on your entire customer list is negligible compared to the cost of the person who used to do it manually.

Judgment work is getting more valuable, not less. The more AI handles the routine, the more your team's time can go toward the work that actually differentiates you. That's not a consolation prize. It's a genuine shift in what your people are for. This maps directly to the two types of problems AI actually solves — and getting clear on which category each task falls into is the first step to knowing what to automate.

The Right Question to Ask About Your Team

The wrong question is: should I replace staff with AI? That question leads to bad decisions and damaged morale, and it's also not how the shift actually works in most service businesses.

The right question is: which tasks on my team's plate are effort problems, and which are judgment problems?

Walk through a typical week and make a list. Every recurring task your team does on a schedule — follow-ups, reminders, reporting, data entry, coordination across systems — that's the effort category. Every task that requires knowing your customers, reading a situation, or making a call that depends on context someone built over years — that's judgment.

The effort tasks are candidates for automation now. The judgment tasks are where your people should be spending more time.

Junior developer job postings have declined 67% over the past 18 months as AI absorbs entry-level work. The same pattern is visible in every industry where repetitive, pattern-based work exists. This doesn't mean those jobs vanish overnight. It means the role of people in those positions has to evolve — away from doing the pattern work and toward supervising, directing, and handling what the pattern can't cover.

A Concrete Example: The 10-Person HVAC Operation

An HVAC company with ten employees — four technicians, two office staff, two sales, and two managers — might look like this on a normal week:

The office staff spends about 12 hours combined on scheduling confirmations, appointment reminders, job status updates to clients, and invoice follow-ups. The sales team spends about 8 hours on initial inquiry responses, estimate follow-ups, and maintenance reminder outreach. The managers spend about 4 hours on review monitoring, job completion documentation, and system data entry.

That's 24 hours a week — three full person-days — on tasks that follow a predictable pattern. An agent handles all of it. The office staff's 12 hours moves to handling escalations, building client relationships, and managing complex scheduling situations the agent can't resolve. The sales team's 8 hours moves to site visits, deeper client consultation, and closing conversations that require real judgment. The managers' 4 hours moves to reviewing outcomes rather than doing data entry.

Nothing about this requires laying anyone off. It requires redefining what each person is responsible for — and being honest about which of their current responsibilities are pattern work versus judgment work.

The Practical Math for Your Business

You probably have someone on your team spending meaningful hours every week on work that follows a predictable pattern. Follow-ups that always say roughly the same thing. Appointments that need the same confirmation process every time. Reports that aggregate the same data in the same format every month. Data that needs to move from one system to another after every transaction.

The intelligence to handle all of that is now available at a cost that's measurably less than what you're paying in labor hours for those tasks. Not in two years. Today.

The businesses moving fastest on this aren't the ones with the biggest AI budgets or the most technical staff. They're the ones that made a clear-eyed list of their effort problems, picked one, automated it, and measured what changed. Then did it again.

Understanding what level of AI your business is actually at is the most honest starting point. Most businesses are at Level 1 or Level 2 — AI is helping individuals with tasks, but it's not running any part of the operation. Getting to Level 3 is where the math starts to compound.

Frequently Asked Questions

How do I bring this up with my team without creating fear? Be direct about what you're doing and why. "We're going to automate the follow-up sequences so you can spend more time on client relationships" is a different conversation than ambiguity about AI "changing the business." The fear comes from uncertainty, not from change itself. The more specific you are about which tasks are being automated and what the freed-up time is for, the more your team can engage with the change rather than dread it.

What happens to team members whose work gets automated? In most service businesses, the bottleneck isn't having too much labor — it's having too much process work crowding out the relationship and judgment work that grows revenue. When the process work is automated, the same people can take on higher-value work. The question to answer first is: what would my team do with 10 more hours a week? If the answer is clear and valuable, the automation conversation is easy. If the answer is "I don't know," that's the conversation to have first.

What's the minimum budget to get started? The tools that handle most small-business effort problems cost between $50 and $300 a month in software subscriptions. The real cost is time — specifically, the time to clearly define what you want the agent to do. Most business owners who do this well report spending 4–8 hours on the initial setup of their first automated workflow. After that, maintenance is minimal.

Is it true that AI is eliminating entry-level jobs? How does that affect hiring? The data on entry-level job postings is clear — they're declining significantly in fields where pattern work is prevalent. For small businesses, this means the pool of people who want to spend their time on data entry and administrative follow-up is shrinking, while the demand for people who can manage, direct, and supervise AI tools is growing. Hiring for adaptability and judgment is more important than hiring for task execution.

How do I measure whether the automation is actually working? Track two numbers: the output the agent produces (follow-ups sent, appointments confirmed, invoices processed) and the downstream outcome that matters to your business (leads converted, client satisfaction scores, on-time payments). If the output numbers are strong but the outcome numbers aren't improving, you have an intent problem — the agent is doing the task but not serving the real goal. Fix the intent, not the volume.

Start With the Math for Your Business

The math on effort automation almost always makes sense faster than business owners expect once the actual cost comparison is laid out. The goal is to get the pattern work off your team's plate — so they spend their hours on the judgment, relationships, and decisions that actually move the business forward.

Associates AI helps businesses identify their highest-cost effort problems and build the automated workflows to solve them — without needing technical staff to do it. If you're ready to run the numbers on your operation, book a call.