Same Discipline, Two Different Companies: How Maybe* Is Building Agent Infrastructure for a PLC and a Startup at the Same Time
By Polly Barnfield, OBE, CEO of Maybe*
We are launching a new series at Maybe* called Scaling Up With Agents. It profiles the founders, operators, and teams we work with as they turn proven domain expertise into production-grade AI Agents. One client per edition. Four posts per edition. Same shape every time.
Most series of this kind launch with a single client. We are launching with two, simultaneously, on purpose. Because the two editions show something the market needs to see right now. The agent-build discipline that works at PLC scale is the same discipline that works at startup scale. The contexts could not be more different. The pattern is the same.
I want to spend this piece explaining what we mean by that, why it matters, and how the two editions illustrate it. Then we will run both editions in parallel over the following four weeks, with one post per edition per week, so you can watch the pattern hold in real time across two very different businesses.
The two clients
Liganova MaSH! is a full service creative agency, specialising in modern storytelling. They produce over 2,500 pieces of content a year for a global luxury automotive brand and reach nearly 100 million people across social. They operate at PLC scale, inside a corporate ecosystem with the standards that come with that. When James Walkinshaw, their MD, decided to deploy AI Agents, getting it wrong was not a learning experience. It was a contractual problem.
groa° is a venture-stage startup founded by Emma Powell. Emma spent five years quietly fixing the retention economics of top-decile ecommerce brands through a methodology she calls Retention-First Growth. The strategy worked. The constraint was capacity. Sustaining top-decile retention performance for a single brand took roughly 70 hours of manual data work per month. The other 97% of the market, the brands who could not afford a specialist agency on retainer, stayed out of reach. groa° is the answer to that problem. It takes Emma's five years of expertise and embeds it into AI Agents that run continuously inside ecommerce brands.
One agency. One software startup. One running on enterprise data scale. One running on venture-stage urgency. We are building both at the same time on the Maybe* platform.
Why launch them together
The instinct, when you have two case studies, is to release them sequentially. Build a story around each one. Give each its own week in the sun.
We are not doing that, because the more interesting story is the comparison.
Most AI thought leadership lives at one extreme of the company-stage spectrum. Either it is enterprise-flavoured (governance, compliance, risk, change management) or startup-flavoured (speed, MVP, product velocity, founder-led iteration). Both flavours are useful. Neither is sufficient on its own. And the market keeps acting as if the disciplines are different.
They are not. After working closely on both builds, I can tell you the underlying decisions are nearly identical. They get framed differently because the company contexts are different, but the architectural choices, the leadership moves, and the team-level outcomes line up to a degree that surprised even me.
Watching the same pattern play out at both ends of the spectrum is the most useful thing we can offer to anyone trying to make AI work in their own business right now. Whether you are a PLC operator wondering whether the startup discipline scales up, or a startup founder wondering whether the enterprise discipline applies before you have a compliance department, this series is for you.
The pattern, in one paragraph
Both clients started with a proven method. Both ran into a capacity ceiling that stopped the method from scaling. Both had to choose between buying workflow automation and building a real agentic system. Both chose the harder option for the same reason: the easy option could not actually solve the problem they had. Both put governance first. Both protected the team from being framed as a cost line. Both ended up with smaller teams doing higher-leverage work, supported by an agent layer that handles the operational load. The vocabulary differs by sector. The pattern does not.
How the four parts will land in both editions
Across the next four weeks, both editions will publish their parts in parallel. Each week takes one theme from the series shape and shows it playing out simultaneously in a startup context and a PLC context. The reader does not have to choose. Both stories arrive together.
Week 2: The Methodology and the Constraint. What does each client actually do, and what is the capacity ceiling that stops it scaling? In groa°'s case, retention work that took 70 hours per brand per month, locking out 97% of the market. In Liganova MaSH!'s case, agency work where administrative tasks were quietly eating billable time and the team was running on triage.
Week 3: The Architectural Choice. Why AI Agents, not workflows. groa° makes the case for continuously observing customer state across five lifecycle orbits rather than firing on static thresholds. Liganova MaSH! makes the case for an orchestration layer behind a single Microsoft Teams chat, not five new tools the team has to learn. Two different architectural answers to the same underlying question: how do you build an AI layer that actually works in real production conditions?
Week 4: A Worked Example. One real scenario, end to end, in each context. groa° walks through an autonomous VIP decay intervention from signal detection to closed-loop learning. Liganova MaSH! walks through a single social brief from inbox to delivery, with five Spud touchpoints along the way.
Week 5: What Changes for the Team. The cognitive partnership angle. What the agents take, what the humans keep. In both cases, the answer is the same shape: agents take the continuous analytical work that humans can never do well, humans get back the work that requires judgement, creativity, and craft.
Why this matters beyond these two clients
Less than 12% of businesses are using AI in a meaningful way. Around 8% have genuinely operationalised it. The gap is not about access to technology. Most organisations have already bought tools. The gap is about how they started.
The 8% started with outcomes, not tools. They involved their teams before building, not after. They made the AI invisible by delivering it inside the surfaces people already work in. They measured success by what their people could do, not by how many people they could remove.
That is true at PLC scale, and it is true at startup scale. It is true for an agency producing 2,500 pieces of content a year for a luxury global automotive brand, and it is true for a venture-stage software company turning a five-year methodology into a product. The discipline scales in both directions.
If you are operating an enterprise and wondering whether the startup speed and discipline applies to you, watch Liganova MaSH! and notice that the rigour comes from the agency's own standards, not from corporate process.
If you are running a startup and wondering whether the enterprise governance applies before you have a compliance team, watch groa° and notice that the governance is what makes the autonomy possible.
The two stories are the same story, told twice, at very different scales. We are launching them together so the comparison is unavoidable.
What to do next
Edition 1 of Scaling Up With Agents profiles groa°. Edition 2 profiles Liganova MaSH!. Both are now live. Both will publish their four parts over the next four weeks, in parallel.
If you want the founder's perspective on groa°, Emma Powell's book Retention-First Growth is the place to start.
If you want James Walkinshaw's perspective on the Liganova MaSH! build, the FutureWeek podcast episode is the place to start.
If you are working on a similar build inside your own business, book a call with the Maybe* team. Some of those calls will turn into Edition 3 of this series.
