Small Business11 minNovember 17, 2026

AI Coding for Small Business: How Gondola Doubled

AI coding for small business in practice: how Gondola doubled revenue with Codex, no new payroll, and which workflows changed first. A step-by-step breakdown.

AI Coding for Small Business: How Gondola Doubled

AI Coding for Small Business: What Gondola Actually Did

Gondola, a platform that helps creatives get credit and get hired, reports that its revenue more than doubled year over year, with no added payroll costs and no significant overhead. Over the same period, its lead engineer increased monthly shipping output by about 50%.

Both numbers trace back to one decision: giving the team access to Codex, OpenAI's coding agent. But the most interesting part of the story is not the engineer who got faster. It is the people who were never supposed to touch the codebase, and what happened to the way ideas travelled from a customer call to a live feature. That chain is where the money moved.

The Gondola case: facts, numbers, and what they do and don't prove

Who Gondola is and why it is a useful test subject

Gondola grew out of a problem its founder and CEO, Jared Kleinstein, noticed while running a creative production agency, Fresh Tape Media. Excellent creative work was being published on social media, yet the people who made it rarely received proper credit. Gondola was built to fix that: it helps creators get credited and get hired by brands.

This is a small team, not a lab. Many of its product ideas come directly from conversations with customers and from operational problems that the business development and leadership people see firsthand. That detail matters. When product ideas originate with the commercial side of the company, the bottleneck is rarely imagination. It is the distance between "we should build this" and "it is live."

The numbers, read carefully

According to OpenAI's published customer story, here is what Gondola reports:

  • Company revenue more than doubled year over year, with no added payroll costs or significant overhead.
  • Brands hired 2.5 times as many creators through the platform as a year earlier.
  • Since adopting Codex in early 2026, the lead engineer's monthly shipping output rose by about 50%, measured in merged changes.

The same story adds that overall engineering output rose by roughly 25% with no change in headcount, and that pull request review time held steady even as the changes themselves grew larger. These figures are company-reported rather than independently audited.

Doubled revenue is the headline. Merged changes and steady review times are the evidence you can actually act on.

A fair reader will ask whether Codex alone doubled revenue. Almost certainly not. Gondola is a small company, and small companies can grow quickly from a low base; the creator-hiring figure shows the market was pulling too. What the case does show is something narrower and more useful for your planning: the company grew without hiring to match, and its product team shipped more of what customers were asking for. That is a capacity story, and capacity is something you can measure in your own business.

What changed first: the processes that moved

The adoption logic is easy to describe and hard to copy carelessly. Three processes shifted before anything else, and none of them was "engineers type faster."

1. The path from idea to working prototype

Gondola's engineers use Codex through the command-line interface and editor tools to speed up implementation, and in GitHub to review pull requests. That is the expected part. The unexpected part is that Kleinstein and Jojo Hesketh, the company's Creator and Client Partnerships Manager, each merged their first-ever code contributions into the main codebase.

Hesketh's work is the clearest example. According to reporting on the case, she sponsored a new offering called Programs and brought a working version to the engineers, who then hardened it. Kleinstein estimated that Codex saved days, if not weeks, and said the feature launched with fewer bugs than hand-coding would have allowed. The old route was a written requirement, a queue, a misunderstanding, a revision. The new one starts with something you can click.

The prototype is the new requirements document. It cannot be misread, because it runs.

2. Timing: building while the customer need is still alive

This summer, Codex let Gondola build and test a World Cup Creator Map while the event was still timely. The team took data the company already held and turned it into a tool that let brands, teams, and agencies find photographers and videographers in the tournament's host cities. Hesketh sponsored the Creator Map and built a fully working prototype with Codex, and engineers polished it for launch.

Anyone who has watched a market window close while a feature sat in a backlog will recognise the value. A seasonal product built three months late is not a late product. It is no product.

3. Daily operations outside the codebase

Codex was not the only tool. Gondola also uses ChatGPT to simplify daily operations such as information retrieval and drafting, which frees up time for product work and for customers. Reporting on the case notes that a person still reviews consequential ChatGPT output. That human checkpoint is part of the design, not a footnote.

Here is how the shift looks in a compact view:

Process Typical starting point What the Gondola case shows
Specifying a feature Written requirement handed to engineering Business owner brings a working prototype
Engineering time Spent on first drafts and routine implementation Spent on hard problems and polishing
Time-sensitive launches Risk of missing the window World Cup Creator Map shipped while timely
Code review Human review, pace set by diff size Larger changes, review time reported steady
Daily admin Manual research and drafting ChatGPT assists, a person checks key output

A step-by-step playbook for AI coding for small business

Gondola's story does not require a large engineering department, which is precisely why it is relevant to owners of small and mid-sized companies. Here is the sequence the case implies, translated into steps you can run in a quarter.

Step 1: Start with the backlog your customers already wrote

Open your list of unbuilt ideas and mark the ones that came directly from customer conversations or recurring operational pain. Gondola's best examples came from that source. Anything that began in a customer's mouth has a built-in test of value.

Step 2: Equip your strongest engineer first

Gondola measured the effect on its lead engineer, not on a committee. Give your most senior developer Codex in the tools they already use: terminal, editor, and the repository host. Track merged changes per month and review time for a baseline before and after. Without a baseline, "we feel faster" is the only evidence you will have, and it will not survive a board meeting.

Step 3: Let the commercial team prototype, with a fence around it

This is the step that separated Gondola's story from a standard developer-productivity anecdote. A partnerships manager and a founder wrote code that reached production. The fence is the engineers: prototypes go to them for polish, and nothing consequential ships on the strength of a non-engineer's draft alone. You are widening who can start things, not who can approve them.

Step 4: Keep the review gate and measure it

Reporting on the case says pull request review time stayed flat while changes grew. That is the number to watch. If review time balloons or bugs climb, you are producing volume, not value. If it holds, your quality control is absorbing the extra speed.

Step 5: Spend the reclaimed time on customers

Gondola's founder framed the result as giving product people more ability to build, engineers more time for hard engineering, and the business team more time with customers, partners, and creatives. He also said that using OpenAI had "increased the ambition of what we think a company our size can build." The reclaimed hours are not a saving to bank; they are the raw material for growth. A team that spends them back on customers is the one that sees revenue move.

If you are weighing similar gains at larger scale, see how Oracle describes turning days of work into minutes with Codex. The pattern repeats: the tool matters less than where you point the freed capacity.

Where AI coding for small business can go wrong

A case study is a sample of one, and the failure modes are predictable.

Attributing everything to the tool. Gondola's revenue doubled alongside a 2.5x rise in creators hired, which suggests demand and execution both did their part. If you promise your board that a coding agent will double revenue, you have set a trap. Promise faster delivery of customer-requested features and measure from there.

Skipping the baseline. Merged changes are an imperfect metric on their own, since a bigger count is not automatically better software. Pair it with review time, defect rate, and, most importantly, a commercial outcome tied to a specific shipped feature.

Letting prototypes become production by default. The engineers' polishing pass is where security, maintainability, and edge cases get handled. Remove it and you will gain speed for a few weeks and then pay for it in rework.

Treating agents as unmanaged staff. As your use grows, you need rules for access, logging, and accountability. Our guide to managing AI agents in the enterprise covers the governance layer that smaller companies tend to add too late. And for a sober look at what agents deliver outside demos, read real-world AI agent performance.

What this means for the way you run the company

The emotional payoff of getting this right is quieter than the revenue figure. When ideas move from customer call to working prototype in days, you stop carrying a mental list of promises the team has not yet delivered. You get the steady sense of control that comes from watching shipped work, review times, and customer outcomes line up on one page, and the relief of not having to hire ahead of every ambition.

The perception shift follows. A board or investor group that sees revenue doubling without matching payroll growth reads that as operational discipline, and it reads the CEO behind it as someone who builds a company that scales on systems rather than on headcount. Your team sees it too: the commercial staff who now ship features become advocates for the change, not casualties of it.

FAQ

Can a small business really use AI coding without a large engineering team? Yes, if you have at least one capable engineer to own quality. Gondola's case shows the model: engineers use Codex to implement faster and to review pull requests, while non-engineers produce prototypes that engineers then refine and merge.

Did Codex alone double Gondola's revenue? No, and Gondola does not claim that. The company reports revenue more than doubled with no added payroll or significant overhead, alongside 2.5 times as many creators hired through the platform. Codex raised the team's capacity to build; market demand and commercial execution did the rest.

What metrics should I track when adopting an AI coding agent? Start with monthly merged changes per engineer, as Gondola did, and pair it with pull request review time and defect rate. Then connect at least one shipped feature to a commercial result, such as pipeline, conversions, or retention, so the gain is visible in money terms.

Is it safe to let non-technical employees write code? It is safe when their work enters the codebase through the same review gate as everyone else's. At Gondola, business-side staff built prototypes while engineers handled the polish before launch, and a person still checks consequential AI-assisted work.

How quickly can a team expect results? Gondola adopted Codex in early 2026 and reported a roughly 50% rise in its lead engineer's monthly output, though results will vary with your codebase and discipline. Run a one-quarter pilot with a baseline rather than expecting a company-wide transformation in weeks.

Which tools did Gondola use? Codex for development, through the command-line interface, editor tools, and GitHub pull request review, and ChatGPT for everyday tasks like information retrieval and drafting.

Gondola's lesson fits on one line: the company that shortens the distance between a customer's problem and a working feature wins the quarter. Your competitors can buy the same tools this week. The only part they cannot copy is the discipline you build around them, and the leaders who build it first will be the ones reporting growth without a matching bill for headcount.

Pick one customer-requested feature, hand it to a prototype, and let the numbers decide whether your company scales on ideas or on payroll.

Have questions? Ask the AI agent right now

Responds in seconds, knows everything about our services and will help with your situation