The CEO of a regulated financial services business asked me how he should scale. He was in the middle of his Series B raise and the board was pushing him on how the next eighteen months were going to look. He had hired well. His product was working. Revenue was moving. He could still tell you everything that was happening in his company because most of it still happened in front of him.

He could not tell me how it all fit together.

That is the question every scale-up CEO is supposed to be able to answer and almost none of them can. Revenue forecasts are easy. Org charts are easy. How the business actually works as a system, what flows into what, who owns each piece, where the work breaks when it breaks, is the question that gets dodged most often and matters most.

The artefact that answers it is a target operating model.

A target operating model is not an org chart. An org chart shows you who reports to whom. An operating model shows you how the work flows. Customer lifecycle on one side, from acquisition through implementation, account management, support, and billing. Product lifecycle in the middle, from build through test, knowledge management, customer training, pricing, and end of life. Supporting functions on the other side, the ones that keep the business running underneath. Management sitting adjacent, driving strategy across all of it.

When I was COO at Lumeon, every component of that map had a named owner, a codified process, a risk score, and a quarterly review. The TOM was not a deck artefact. It was the source of truth for how the company actually worked. New hires found out who owned what by looking at it. Existing staff improved their own working lives by editing and improving the process they ran. The board read the colours on it when they wanted to know where the risks were.

That was useful in 2022. It is essential in 2026.

Every business is being told to deploy AI. Almost no business is being told what AI needs to deploy into.

An AI agent cannot route a sales-to-success handoff if nobody has ever drawn the handoff. An LLM cannot triage a customer escalation if there is no map of how escalations got triaged before. An automated workflow cannot reduce the cost of onboarding if the onboarding flow lives in three different people’s heads. The agent does not infer the shape of your business from the shape of your data. It needs to be told. The TOM is what tells it.

This is the inversion most operators have not caught up with yet.

The TOM used to be a back-office document. Useful, but quiet. Now it is the substrate AI lands on.

If you can describe a process precisely enough to write a codified swim lane diagram, you can describe it precisely enough to put an agent inside it. If you cannot, you cannot. The work of building a clean operating model and the work of preparing the business for AI are the same work. They were always the same work. AI just made the cost of not doing it visible.

The CEO I spoke to in financial services did not need an AI strategy. He needed a target operating model. Once he had one, the AI strategy would write itself, because he would know which boxes had narrow enough scope and clean enough inputs that an agent could take them, and which boxes needed human judgment and would for the foreseeable future. Without the map he was guessing. With the map he was deciding.

The businesses that win the next decade will not be the ones with the biggest AI budget. They will be the ones whose operating model is legible enough for AI to actually do work inside it.

If you cannot describe the business, no model can save it.

When a growth-stage SaaS business needs to redesign its commercial motion, the work has to happen alongside the business, not in place of it. The team is still selling. The CEO is still running. The board is still meeting. Clarity has to arrive without the company pausing.

Today I’m launching Ortent Advisory to do that work in flight.

Ortent is an advisory practice for growth-stage SaaS businesses building in Life Sciences, Digital Health and AI. Three engagement shapes. Each with a defined scope, fee and timeline.

Course Correction Sprint. Four to eight weeks. Fixed fee. One commercial question answered cleanly. Pipeline reset, operating rhythm, AI-era go-to-market, partner strategy or alliance design. You walk away with a defensible answer and a plan the leadership team can run on Monday.

Fractional CGO or COO. The embedded operator. Two to four days a month for six to twelve months. Revenue, partnerships and alliances on one side. Revenue operations, forecast discipline, partner ops, and ISO 9001, ISO 27001 and SOC 2 readiness on the other. The brief is to build something durable and hand it on to the permanent team.

Non-Executive Director. The long view. Quarterly board, plus chair and CEO 1:1s, light committee work, named introductions, exit-prep authority. Best where commercial clarity, operating maturity and partner-led growth are material to value.

Best fit is five-to-thirty-million ARR. SaaS businesses serving Life Sciences, Digital Health or AI buyers. Particularly where enterprise or regulated customers are the next growth unlock.

Four prior exits as a commercial operator taught me a simple lesson. Companies are rarely undone by strategy. They are undone by how revenue, operating discipline and partnerships fit together under pressure.

AI is now reshaping how customers buy in regulated and scientific markets. It is also reshaping how growth-stage SaaS teams are expected to operate: which tools to use, when, how to operationalise them. The businesses that get this right build coalitions early. The ones that don’t fall behind.

The first three Ortent field guides are live at ortent.co. Field notes ship every two weeks at ortent.substack.com.

If you have a board seat, a fractional mandate or a commercial reset coming up in the next ninety days, get in touch via ortent.co/contact. Thirty minutes to see whether there’s a fit.

Andrew Wyatt
Founder, Ortent Advisory Ltd
ortent.co

The auditor asks what your model did at 14:32 on a Tuesday in March, on a mortgage application that was declined and is now sitting in front of the Financial Ombudsman. You either pull up the run, show her which component handled what, and walk her through the decision. Or you don’t have a system. You have a demo.

That moment, repeated across every regulated industry I have worked in, separates AI products from AI press releases.

Most boards still ask which large language model to bet on. Wrong question. One big model is a single point of failure. Opaque. Expensive. Brittle on the narrow tasks that actually matter inside a business. The accuracy headlines are written against general benchmarks. Your benchmarks are not general. Your benchmarks are: did the system route this clinical pathway correctly, did it triage this support ticket, did it size this account, did it flag this contract clause. On those, a small domain-tuned model usually beats the headline LLM. A chain of small specialised components beats either of them.

The expensive version of this mistake is taking one frontier model and training it on your entire business. Every contract, every ticket, every Slack channel, every customer record, fed into a single fine-tune. The pitch sounds great. The bill is huge, the system cannot be audited the way any regulator wants, and you are now locked into one vendor’s roadmap, pricing, and licence terms. When any of those change, and they will, the cost of unwinding is a board-level event.

AT&T have just put numbers behind this in public. They rebuilt “Ask AT&T” away from heavy general-purpose models and toward a mix of smaller, domain-specific ones for the bulk of their twenty-seven billion daily tokens. The reported saving is around ninety percent. They run agents that plan and execute narrow tasks. They back open-source telco-specific models because generic LLMs were not built for telecom data. They keep humans in the loop on anything that touches a customer or the network. Hybrid architecture. Multiple models. Best fit per task. That is not an AI strategy lifted from a vendor deck. It is what an operator does when the bill arrives.

The same shape shows up wherever the work is regulated and specialised. When I was at Sapio Sciences, SigmaticOS was built as an orchestration platform, not a single model. Over a hundred agents, each tuned to a part of the drug discovery loop, sat under an orchestrator. The orchestrator picked the right agent, handed it the right context, then folded the result into the next step. In silico work on one side, lab automation on the other, model training in between. The LLM was not the product. The orchestration was the product.

Lumeon proved the same point without using AI at all. A care pathway is not one decision. It is fifty, taken across clinicians, systems and patients over weeks. The platform routed each step to the right component: the rules engine, the messaging layer, the EHR integration, the patient-facing app. Deterministic by design, because clinical care has to be. Every step inspectable. Every decision traceable to its source.

That is the bar AI deployments in regulated domains have to clear. Probabilistic components can sit inside this kind of architecture. They cannot become the architecture.

Here is the part most teams still miss.

The architecture above is necessary. It is not enough. What makes it actually work inside a regulated business, what makes it survive a board paper, a SOC 2 audit, an FCA inspection, a customer escalation, is the audit trail. Every call to every component, every input, every output, the version of the model, the version of the prompt, the data the agent saw. Replayable. Searchable. Owned by you, not the vendor.

Suresh Rajashekaraiah and others have started naming this layer as a category in its own right: observability for agentic AI. Agentic systems do not fail the way traditional software fails. The failures are not stack traces. They are emergent and probabilistic.

You cannot debug what you cannot see. You cannot govern what you cannot replay.

Boards that have lived through a serious incident understand this in their bones. Boards that have not are still treating observability as a cost line.

The pattern is not complicated. It is just unfashionable. Stop asking which frontier model to license. Ask three questions instead.

Which tasks are narrow enough that a smaller, domain-tuned component would beat a general LLM. The honest answer is: most of them.

What does the orchestration layer look like, and who owns it. If the answer is “the model vendor,” your moat is rented.

What is the audit trail, where does it live, and can a non-engineer reconstruct any single decision the system made in the last twelve months. If the answer is no, you do not have a regulated AI product. You have a liability waiting for a Tuesday in March.

The companies that win the next phase of enterprise AI will not be the ones with the biggest model. They will be the ones who can show their working.

When I was in my twenties I played elite volleyball. As elite as volleyball gets in the UK, which is not very.

One thing from those courts has shaped every hiring decision I have made since.

Volleyball is a real team sport. The opposition spends the first three points working out who your weakest player is. Once they have found him, they serve on him every time. They hit through him. The match is decided not by your strongest player but by how exposed your weakest one is.

No two players on the team had the same strengths. The job was to harness the differences, not smooth them out.

John Nash was the England head coach at the time. He said something I have repeated in interviews and board rooms ever since.

Most leaders do the opposite. They hire people they like. People who went to similar universities. People who tell similar jokes. People they would go for a beer with after work. The team feels comfortable in the room and looks one-dimensional everywhere else. The opposition finds the weak point and there is no one with a different strength to cover it.

I have been guilty of this. The Clearswift hire for head of inside sales is the moment I remember most clearly.

Inside sales people spend their day on the phone. They get rejected ninety-eight percent of the time and dial again. They are loud. They run competitions. They buy donuts and announce that the team can have one when they have hit ten grand in the morning. They are a very different breed to me.

I interviewed five or six candidates. None of them felt right. Then I realised why. They were not like me. I was looking for someone I would enjoy a glass of red with. The right hire was someone who would buy the team a tray of donuts and run a leaderboard on the wall.

We hired her. Bubbly. Gregarious. Competition every morning, donuts on Fridays, leaderboard on the wall. The phones lit up. The pipeline grew. I went for beers with her in the end. That was the consequence of the hire, not the reason for it.

Apertio was the same lesson on a different role.

I had inherited a likeable product manager. Everyone got on with him. He said yes to everyone and built nothing. The roadmap drifted because no one could remember what had been promised in which meeting.

I replaced him with an ex-army officer. Different culture. Different background. Different way of running a room. He committed to a plan, shared it, and held the line. People knew where they stood. They did not mind hearing no, because the yeses got delivered six months later when the release shipped.

Both hires would have failed a “beer with this person” test on the day. Both turned out to be the most consequential hires I made in those companies.

The same pattern shows up on boards. The chair or the lead investor nominates someone they worked with at a previous company. The CEO knows the candidate is a known quantity and struggles to argue against it. The board ends up with another version of the people already in the room. The one new perspective the company actually needed gets quietly priced out of the search.

The test for any leader is simple.

Who on your team is better than you at the thing you are worst at?

If the answer is no one, you have not been hiring. You have been recruiting yourself.

Ask your investor directors how they would measure a software development team.

Watch what happens.

Most cannot answer. They are smart people. They are also wrong about your business, and they don't know it yet.

Most VC and PE-backed boards are full of investors. Numbers people. They will help you build a budget, manage burn, run a refinancing, read a model, and ask the right questions when revenue slips. That work matters. But it is not the work that decides whether the company wins.

Channels to market. Positioning. Use cases. Where to sell first. How to sequence the wins that unlock the next ones. None of that lives in the model. All of it lives in the day-to-day of running a software business. On the boards I have sat on or watched closely, this is where investor directors go quiet.

If you are lucky, your investors at least know your market. They have looked at twenty companies in your sector. They know the buyer and the budget cycle. That helps. But knowing the market is not the same as knowing how to run a software business inside it. From the outside, those skills look similar. In the boardroom, they are not.

The worst case is the investor who fills the silence. They have spoken to their colleagues. None of them could quite get the proposition straight either. So they decide the answer is to tell the CEO how to position the company. Where to focus. Which customers to chase. The advice is delivered with confidence. It lands in the business like a wrench in the gearbox. Two quarters later the plan is off and the board cannot say why.

Apertio is the case where this did not happen.

Our investors were not just numbers people. They were software people. Some had built and exited their own companies. One was strategic. They could see why what we were building would matter to them in a few years.

They argued with the CEO about things that actually moved the company. Use case. Pricing. Channels. Geography. Which customer wins to chase first to unlock the next ones. The audit committee did its job. The model got read. The budget got tightened when it needed to be. But the operating debate happened on operating ground, in the same language the management team were already thinking in.

The Motorola part of that story is a complication. That is for another post. A strategic investor on the cap table can scare other buyers off. The cost has to be weighed against what they bring while you are still building. On the Apertio board, the operating instinct was the part that mattered most while we were still building.

Lumeon got there eventually. The board went through several mixes before it found the right balance of investor discipline and software operating sense. That cost time. Time on a board is expensive.

Paragon was different again. The CEO held fifty-one percent. He could push the agenda through without friction. The board became a governance and reporting sounding board. Useful for what it was. It was not a board that pushed back. There is a version of that company that exits earlier and for more, with a different cap table and a board that argues.

The hire that makes the biggest difference is the operator NED with both market and software experience, in the geographies you are trying to win.

Not market alone. Not software alone. Not someone who has read about your sector. Someone who has run a software business inside it, in the place you are trying to win it.

That is the hire that changes how a board reads the papers. They notice what is not on the slide. They argue with the CEO in the language the CEO is already thinking in. They give the investors something to push against that is not just a louder version of what the investors already think.

Most boards do not have that hire. Most chairs have not asked the question that would tell them they need it.

Ask your investor directors how they would measure a software development team.

Healthcare is not like any other market tech sells into.

Most founders assume the hard part is regulation. HIPAA, GDPR, data privacy, long security reviews. Those matter. But they are not what kills most healthtech deals.

The real killer is change management. And most startups ignore it completely.

They build great tech. They assume every clinician will want it. Then they hit the reality of how health systems actually work.

Health systems are huge. Staff rotates constantly. Clinical workflow is already overloaded. Getting anyone to adopt a new tool is a full program of work, not a rollout slide.

My best illustration of this came from New York Health and Hospitals. It is the largest public health system in the United States. You probably know the building. It is where the TV series New Amsterdam was filmed.

At Lumeon we had deployed a digital solution for NYHH. It managed a specific cohort of diabetic patients. The mechanic was simple. We sent each patient an SMS asking for their blood glucose reading. If the system spotted a downward trend, it automatically added the patient to a call list for clinical follow-up and a medication review.

Before our solution, a clinician had to phone every patient. Take the reading over the line. Work out on the spot whether anything needed to change.

I went to see the lead clinician at NYHH. She was delighted. She said she could not understand how they had coped before. The benefit was obvious. Workload went down. Clinicians focused on the patients who actually needed them.

Job done.

Except it wasn’t.

I asked her if there were any problems. Yes, she said. We have to tell every consultant in the building that this service exists. They need to know how to enrol their patients onto it.

Fine, I said. We can send an email. Put up some posters.

She laughed.

“Do you have any idea of the churn rate in this building? Staff rotate every month. By the time I have trained someone how to enrol a patient, they have rotated out. And besides, take a look at the board behind me.”

I turned around.

The wall was covered. Posters, sticky notes, flyers, printouts. Every other program and service clinicians were meant to be enrolling patients onto. Dozens of them. Our service would be one more sticky note in the pile.

That is the lesson most healthtech founders learn the hard way.

You can have the best clinical tech on the market. But if you cannot solve for staff rotation, consultant attention, and the noise inside a busy health system, your product is dead on arrival.

Most founder CEOs do not fail because they cannot solve problems. They fail because they cannot stop solving them.

The trait that makes a founder is the same trait that eventually breaks the company. They love problems. Any problem. They do not particularly care whether it is the one the business needs solved. New problems are interesting. Old problems, even the ones the company is paid to solve, become dull. Repetition bores them. And the moment a brighter, shinier problem appears, they are off, and the company goes with them.

I have watched this pattern play out across four decades.

At Intuwave, I was brought in after this had already happened. The founding team had built a capable mobile platform and tried to sell it in a dozen directions at once. Each new conversation with a new prospect opened a new idea. Each new idea added a new feature. The product sprawled. The pipeline did not. The turnaround was not technical. It was behavioural. We picked one problem, remote device support for mobile operators, and ignored everything else until we had proof points. The technology had been capable the whole time. What changed was the refusal to be distracted.

At Apertio, the founder had the opposite instinct. We replaced legacy home location register infrastructure in mobile networks, and we did nothing else until we were unquestionably the best at that one thing. It was not glamorous. It was not varied. It required saying no to adjacent opportunities for years. That discipline is why the company scaled, and why Nokia bought it. Founders like that are rare.

Most are not wired that way. They see the market they are winning and assume it is solved. They see the market next door and assume it is interesting. They confuse motion with progress. They mistake the dopamine of a new problem for strategic insight. And because they are the CEO, the whole company has to follow them into whatever they find interesting that quarter. Product teams pivot. Sales pitches mutate. Customer success loses the plot. Partners stop understanding what the company actually does.

This is why so many great founder CEOs become mediocre operators. The skills that got the company from zero to one are not the skills that get it from ten to a hundred. The company needs repetition, proof, references, land and expand. The founder needs novelty. At some point, those two needs diverge, and the business starts to pay for the mismatch.

The founders who scale are the ones who recognise this in themselves. They either impose external discipline, through a strong COO, a demanding board, or investor pressure, or they hand the operating role to someone else and move into a chair or product visionary seat where their instinct is an asset rather than a liability. The ones who do not tend to burn through cash, confuse the team, and watch a better-focused competitor take the market while they chase the next interesting thing.

If you are a founder reading this, the test is simple. Look at your last four quarters. How many different problems did you ask the company to solve? How many of them were the one you are actually paid to dominate? If the answer makes you uncomfortable, that is the work.

Most platform companies don’t fail because the technology is weak. They fail because they try to do too much, too early, for too many people.

I’ve seen this pattern repeat across multiple companies.

At Lotus, Notes eventually became incredibly sticky. Entire organisations ran critical workflows on it. But it took a long time to get there. Early on, it tried to be everything: database, email, applications. It struggled to land. What worked wasn’t the platform story. It was very specific use cases, document workflows, approvals, regulatory processes, where partners could build practical applications that solved real problems. That’s where adoption came from.

At Intuwave, I was brought in to fix the classic mistake. The team had built a mobile platform and tried to sell it as a platform. It went nowhere. The turnaround came from focusing on one tangible problem: helping mobile operators support devices remotely. Suddenly there was a buyer, a budget, and urgency. Same technology. Different outcome.

Apertio got it right from the start. We focused relentlessly on one problem: replacing legacy home location register infrastructure in mobile networks. It was a massive, expensive, visible pain point. We solved it better than anyone else. That focus drove scale, credibility, and ultimately an acquisition. The broader platform capabilities came later, once we had earned the right.

At Lumeon, I watched this tension up close. The platform orchestrates complex healthcare workflows at scale across patients, clinicians and systems. The harder question is the one every platform company eventually faces: which single use case to own first. That decision is what separates a capable platform from a category winner. It’s also one of the hardest calls a founder or CEO ever makes.

This is the uncomfortable truth. Focus feels like constraint, especially to founders. Most entrepreneurial CEOs are wired to solve problems. New problems are interesting. Repeating the same solution in the same market is not. So they add features, expand use cases, chase adjacent opportunities too early. That instinct is often what kills the business.

The discipline required to win a market is boring. It’s repetition. It’s saying no. It’s doubling down on the same problem until you are unquestionably the best at solving it. Only then does the platform matter.

The pattern is simple. Pick a target market where the pain is real and urgent. Build a solution that clearly beats alternatives. Drive adoption until you have momentum and credibility. Then expand carefully into adjacent areas.

Do it in any other order, and you’re just another platform looking for a problem.

You can tell how a company is going to exit months, sometimes years, before it happens.

In the companies that exit well, people are still arguing about the roadmap. They are still taking customer complaints personally. Someone is still quietly pushing back on a bad deal because the margin structure is wrong. The business is being run.

In the companies that don’t, the arguments have stopped. The conversations have tightened. People are waiting. You can feel the shift the moment leadership decides the company is an asset to be sold rather than a business to be built.

I’ve been through four exits across different markets and decades. What I’m more certain of now than I was ten years ago: the companies that exit best are the ones that stopped thinking about the exit first.

One of them is my clearest cautionary tale. A business had been sold to a large security player at the peak of the dot-com boom. When the bubble burst, the original investors bought the assets back and put a new company together. The thesis was a quick onsale. It didn’t happen, and it didn’t happen because nobody was running a company. They were running a sale process that had no buyer.

The signs were everywhere. No investment in the organisation. No investment in IT. No investment in anything operational. Customer satisfaction was low because customers were an afterthought. Why invest in the relationship if we were leaving in eighteen months? At one point the company sued a major customer for IP infringement because the customer was under-licensed. A good operator would have walked in, fixed the licensing, and turned it into a larger contract. The business took the million and the lawsuit, because a million in the bank would look better in the data room than a healthy long-term customer.

That is what a company run for exit actually looks like. Short-term cash over long-term relationships. Feature freezes dressed up as focus. Channel partners neglected because they don’t show up in this quarter’s numbers. And underneath all of it, a culture where good people start to behave badly because the signal from the top is that nothing matters except the transaction.

It took a new team, a reset of the operating model, and several years of rebuilding before the investors got their money back. Not the quick exit. The long one.

Apertio was the opposite case. We built a product, established a new market category, and ran the company as if we were going to own it forever. We were deep into exploring IPO options because the plan was to stay independent.

At Mobile World Congress one year, a journalist was interviewing me about the Apertio story and clearly didn’t believe any of it. He thought I was making it up. I paused and asked him what would convince him. He said a customer telling him it was true. I asked him to wait, walked into the next meeting room, and came back in with the Global CTO of T-Mobile. “This do?” I asked.

That was the company we had built. Customers who would walk across a conference to vouch for us. Tier 1 telcos who picked up the phone. A product so obviously winning in the market that a journalist’s disbelief was the most honest reaction we got that week.

Nokia came to us because their commercial teams could tell, in real time, that they were losing every deal where Apertio showed up. The acquisition was defensive. They bought us because they couldn’t beat us, and because an IPO would have put us permanently out of reach. We weren’t positioning for an exit. We were positioning to win. Apertio went on to become the core of Nokia’s network business.

That is the pattern that actually repeats. Not a clean data room. Not a rehearsed equity story. A company so clearly winning that a buyer concludes it is cheaper to own you than to compete with you.

The practical version of this for PE sponsors and board chairs is a question, not a checklist. Are you pushing your portfolio company to build the business, or to stage it? One of those produces exits. The other produces lawsuits, churn, and write-downs dressed up as strategic decisions.

You can usually tell which one is happening by listening to what the employees are arguing about.