Artemis Petconnect Insights · Everything Begins with Understanding · Paper I / How we work
The New Economics of Complexity
How a four-person company uses AI to coordinate work that once needed agencies, freelancers and departments

Artemis Petconnect Insights · Everything Begins with Understanding · Paper I / How we work
How a four-person company uses AI to coordinate work that once needed agencies, freelancers and departments

For our October 1 relaunch, we rebuilt both of our websites at once: the company site and the site for our product, askKynos™. Earlier in the year, no AI assistant we tried could hold that much work at one time. By the fall, Claude could work on both sites in the same conversation and keep them saying the same thing. We went build after build, version 5 through version 12 and then a final. Along the way we caught a page promising pet sitters they would be “live” within six hours. The truth was that they would be approved within six hours. Going live took a few more steps.
That moment is this paper in miniature. AI did not remove the complexity of building a business. It changed what a small, experienced team can coordinate: work that once needed agencies, freelancers and separate departments can now be explored, drafted, challenged and revised inside our own decision process. But the AI did not decide what we could promise. We did. The value of AI depends less on how much it produces than on the system of review, context and accountability around it.
The economics are real, and they sit on the cost side. We incorporated on April 6, 2026. By November 16, four of us, with no employees and no outside funding, working nights and weekends, had built an app on iOS and Android offering five kinds of pet care, two websites, a Vet Assistant and an advertising program. The AI tools the four of us work with cost about $155 a month in total. Hiring even two people to do the work we handed to AI would have cost about $120,000 in wages over the same seven months. What AI changed is the price of attempting complexity. Whether that attempt earns a return is the market test that began on November 16. Along the way we describe three kinds of leverage AI gives a small team (exploration, coordination and judgment) and six practices any organization can apply.
Consider a new service launch. Pricing must be defined. Service names must match what can actually be delivered. A website must explain the difference between similar offerings. Advertising must make promises consistent with the product, and each advertisement must point to the right page. Legal, operational and customer-experience implications must be checked. Change one decision and work elsewhere becomes wrong.
For a small business, complexity is rarely one hard task. It is an accumulation of dependencies. A landing page can be well written and still be wrong because its service definition is out of date. A campaign can be creative and still fail because it implies a relationship with caregivers the company never intended. The six-hour promise was exactly this kind of error: a well-written sentence that no longer matched the product. Faster drafting fixes none of these problems.

The traditional answer is specialization: agencies, freelancers, departments and handoffs between them. Those resources remain valuable, especially for regulated, technical and high-stakes work. But every handoff adds cost and a chance to lose context. AI offers a different option: support for thinking, drafting and checking that lives inside the team’s own decision process.
We did not begin with a theory of AI and organizational design. We are four co-founders, one family, building a pet-care business on nights and weekends. The CEO & President sets the business direction and writes the product specifications, the websites and the company’s stories. The CIO & VP Engineering designed the architecture and puts the product together, and the CFO & CTO writes much of its code and leads the data work. The CMO & VP Communications owns the brand, the interface and the experience of Kynos, the companion inside askKynos.
Each of us works alongside AI, and different systems do different jobs: Claude for website content, revisions and much of the product work; ChatGPT for research, positioning alternatives, campaign critique and cross-checks. Every decision comes back to one of us.
There is a difference between asking one AI for a finished answer and coordinating several rounds of work with explicit review. One tool’s output becomes another’s input. A contradiction sends us back to the source. When the systems disagree, we don’t take a vote; we go back to the product decision and check the facts.

In June, before we launched, our CMO pushed back on how our first website talked to sitters. It led with a bare percentage and talked about what sitters could “earn,” which read like a job advertisement. We were writing to independent pet-care business owners. We used AI to test alternative framings against the relationship we wanted, and the line became “keep more of what you earn,” with the 85% underneath as proof. Later ads carried it forward as “Keep More of What You Make.”
That was not copyediting. It required deciding what business relationship we wanted, then making the language, the service descriptions and the ad destinations agree with that decision. We recorded the approved framing and the reason we rejected the old one, so later work would not bring it back.
The same thing happened with boarding and house sitting. Treating them as interchangeable would have produced misleading descriptions and links. When a landing page seemed to be missing, we went back to our build sheet instead of deleting a service. Each correction was small. Together they show why a record of decisions matters more than any single draft.
Generative AI writes with remarkable fluency. Fluency is not a warranty. AI cannot establish that a fee promise is approved, that a page is deployed, that an offering is lawful everywhere it appears, or that customers want it. A workable model needs a clear line between assistance and responsibility.

Decades in enterprise technology teach that the valuable question is not “Can AI write the page?” but “What is the customer being promised, and can we stand behind it?” Our line follows from that. AI can propose, reorganize, compare and challenge. We validate claims, approve product behavior, check the consequences for pet parents, caregivers and pets, decide when something is ready and own the result. In a pet-care company, safety sits on the human side of that line without exception. An unverified answer does not become reliable because a second AI repeats it.
This can look conservative. In practice it protects the advantage AI creates. If fast production multiplies the material a business must later correct, the gain disappears. Verification is not the opposite of speed; it is what makes speed usable.
Verification is not the opposite of speed; it is what makes speed usable.
Much of the public conversation about AI is about prompts. In sustained business work, the discipline that matters is the loop: frame the problem, generate options, critique the claims, verify the facts, then decide. Each stage has its own test. When a check fails, we go back to the requirement instead of polishing a wrong premise.

This is why context is an asset. Each of us works in an AI project that holds our role: the CMO’s knows the interface, the engineers’ know the app, and the CEO’s knows the specs, the sites and the story. A decision made yesterday is still there in the morning. Without that, an assistant can produce persuasive material that is already out of date. As we put it to ourselves: logging in is not the same as being known.
There are four of us, no employees, and a Claude Team plan. That plan is our headcount: five seats at $25 each, one for each of us plus a fifth that goes to our CTO, because writing code burns through usage fastest. Add a ChatGPT subscription for research and critique and a Grok subscription for a second opinion, and the AI tools the four of us work with cost about $155 a month. When a seat hits its usage limit, the assistant makes us wait until it resets. We call it jail.
There are four of us, no employees, and a Claude Team plan. That plan is our headcount.
A couple of nights before the October 1 launch, our CMO went to jail for the first time, in the middle of the work. She sent a frantic note: “Help, I ran out of spend in my account.” Our CEO’s first instinct, after decades in enterprise software, was to give her whatever she needed, right away. The rest of the team stayed calm, and she was soon back at work. At the launch meeting we asked what she had needed. “Nothing,” she said. “I’m good.”
That exchange captures the new economics. The limit is the budget. Nobody has ever turned jail into a request to hire someone, and the work still got done. Seven months earlier, attempting this much connected work would have meant hiring, contracting or raising money first. We did none of those.
We apply the same discipline inside the product. The AI that runs inside askKynos is a separate operating cost that grows with use, so smaller, faster models handle simple jobs, such as Kynos’s short greetings, and larger models are reserved for work that needs deeper reasoning. We chose that deliberately, because we fund the company ourselves.
Without AI, we would not have done all of this work ourselves. We would have hired people to do it, and spent our own hours directing and critiquing their work. Count the roles the four of us cover with AI’s help, and the list is long:
| Co-founder | Roles covered, with AI’s help |
|---|---|
| CEO & President | Product specifications · web design · web development · copywriting and editing · press and media · first drafts of legal documents for counsel to review |
| CMO & VP Communications | Brand design · illustration · interface design · interface changes, made herself |
| CFO & CTO | Application code · AI features · data science · finance: business banking, company cards and paying the bills |
| CIO & VP Engineering | Architecture · putting the whole product together · back-end, payments and integrations engineering · the push notification system · releasing the apps and both websites |
That is well over a dozen roles across four people.
Our own working estimate is that AI lets each of us do the work of about three people. We can’t prove that number, so here is a smaller one. Suppose we had hired just two full-time people since April, say a web developer-designer and a marketing writer, at about $48 an hour, the US median for web developers and digital designers [3]. Over the 32 weeks from incorporation to launch, that comes to about $120,000 in wages alone, before benefits, recruiting or management time. Our AI tools over the same period cost about $1,100. Two hires is a conservative count; the real comparison is larger.
A low cost base is not only cheaper. It changes what a company can choose to do. A platform that carries salaries and answers to investors has to earn from every booking. We don’t, at least not yet. What we need is density: enough pet parents and sitters in each community for the marketplace to work. So from November 16, sitters keep 100% of every booking while we grow, with no platform fee. We are trading near-term revenue for local density, and our cost structure is what lets us make that trade.
A second choice follows the same logic. askKynos is non-exclusive: sitters keep their own clients and can keep working on other platforms. We don’t believe in lock-in. We believe in a better match. Our principle is simple: don’t trap the relationship; make the relationship better.
None of this makes the work free, and none of it is revenue. The full cost includes our time, engineering, infrastructure, testing, marketing, operational systems, and the effort of reviewing and correcting what AI produces. The new economics are better understood as three kinds of leverage:
Outside research supports careful interest rather than sweeping claims. In a randomized field experiment with 776 Procter & Gamble professionals, individuals working with AI matched the performance of teams working without it, and AI helped people produce solutions that crossed functional silos [1]. A separate study of customer-support agents found a 15% average increase in issues resolved per hour after an AI assistant was introduced, with very different effects across workers [2]. Both studies involve large organizations; neither establishes a comparable effect for us.

Before November 16, we reconciled service definitions, website copy and advertising for five kinds of care: pet daycare, boarding, drop-in visits, house sitting and dog walking. That shows organized preparation. It does not show adoption.
The measures that matter come next: completed pet profiles, activated caregivers, registrations by area, booking requests, completed bookings, repeat use, customer feedback and real acquisition costs. Tracking our own hours, revision cycles, errors caught and outside spend would also help judge the working model itself. Where those numbers do not exist yet, we will not fill the gap with estimates presented as facts.
Each practice below is something we learned by doing it, or by getting it wrong first.
None of this works without people who know the domain well enough to spot a wrong answer, and who accept the friction of checking. A four-person company is still a four-person company. It does not gain an engineering department, an agency or a customer base because it pays for AI. What changes is the range of connected work four people can coordinate, through exploration, coordination and judgment leverage.
We didn’t add AI to a pet-care marketplace. We built a pet-care marketplace around what AI makes possible. That claim now has to be tested against real business results, not celebrated before customers have spoken.
Our organizing idea is simple: everything begins with understanding. It applies first to understanding the business problem and how people and AI work together. The same idea shapes how askKynos treats each individual pet (Paper II) and how applications can make their capabilities, context and limits understood (Paper III). The future of small-company work may not belong to whoever produces the most AI-generated material. It may belong to whoever knows what matters, tests its assumptions and turns fast exploration into reliable decisions.
Artemis Petconnect, Inc. is the company behind askKynos™, a pet-care platform built on the belief that every pet deserves to be understood. Four equal co-founders, one family, founded the company and built askKynos on nights and weekends: the CEO & President; the CFO & CTO, a PhD data scientist; the CIO & VP Engineering; and the CMO & VP Communications. askKynos offers five kinds of care (pet daycare, boarding, drop-in visits, house sitting and dog walking) built around one persistent story for each pet.
Company: https://www.artemispetconnect.com · Product: https://www.askkynos.ai
The stories in this paper are told in more detail in our founder series, Eating Our Own Dog Food:
[1] Dell’Acqua, F., Ayoubi, C., Lifshitz, H., Sadun, R., Mollick, E., Mollick, L., Han, Y., Goldman, J., Nair, H., Taub, S., & Lakhani, K. (2025). The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise (NBER Working Paper 33641). National Bureau of Economic Research. https://www.nber.org/papers/w33641
[2] Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at Work. The Quarterly Journal of Economics, 140(2), 889–942. https://doi.org/10.1093/qje/qjae044
[3] U.S. Bureau of Labor Statistics. Web developers and digital designers. Occupational Outlook Handbook. Median hourly pay of $47.85, May 2025. https://www.bls.gov/ooh/computer-and-information-technology/web-developers.htm
Author: Artemis Petconnect, Inc. · Title: The New Economics of Complexity · Series: Artemis Petconnect Insights, Everything Begins with Understanding · Paper: I · Published: November 2026 · Address: https://www.artemispetconnect.com/insights/the-new-economics-of-complexity
License: © 2026 Artemis Petconnect, Inc. This paper is licensed under the Creative Commons Attribution-NoDerivatives 4.0 International License (CC BY-ND 4.0): you may share and republish it unchanged, including commercially, with credit to Artemis Petconnect, Inc. https://creativecommons.org/licenses/by-nd/4.0/ The Artemis Petconnect and askKynos names and logos are not covered by this license. ChatGPT, Claude, Grok and Rive are trademarks of their respective owners; no partnership is implied. Figures are conceptual illustrations. Cover illustration created with ChatGPT and edited by Artemis Petconnect.
How to cite
Artemis Petconnect, Inc. (2026, November). The New Economics of Complexity. Artemis Petconnect Insights, White Paper I. https://www.artemispetconnect.com/insights/the-new-economics-of-complexity