Tom Binza

My CV has the dates, the titles and the numbers. This page is the part a CV cannot carry: how I actually work, what I have taught myself, and what I believe about the job.

The pattern that runs through all of it is the same. Statutory consenting on a railway, where eleven technical disciplines had to agree before anything could be submitted. Two companies merging, where nobody could say which team owned what. A support queue where the real problem was never the ticket in front of you. In each case the work was not heroics. It was making the system legible first, then fixing the part that was actually broken.

I am drawn to that in my spare time too, which is probably the tell. I build things to find out how they work, and I read almost exclusively about ordinary people solving what an institution got wrong.

I direct a small team of AI agents, and it has changed what I think one person can run.

I am not a developer. I am a technically literate operator who uses AI to design, build and maintain working software. I am learning to read code, I write the prompts that produce it, and I debug what comes back.

The setup is a roster of named Claude Code sub-agents, each with a defined remit, its own profile and a persistent memory file. I direct, they execute. It runs through Telegram so I can give instructions from my phone, and a written brief lands in a folder at 8am each morning.

pm / engineering

specs, builds, ships

design

interface and identity

strategy

positioning, sequencing

legal

terms, policy, copy risk

finance

models and unit economics

marketing

launch and messaging

auditor

catches it before it ships

The auditor is the one I would defend hardest. A pre-build reviewer whose only job is to find the mistake before it reaches anything real is the closest thing this way of working has to a safety rail.

A complete social app, built end to end

Product spec through to deployed code, legal copy and a financial model. It has been through a full technical audit, the company and the developer accounts are set up, and it launches shortly. I did the lot: the build, the terms, the store submissions, the numbers. Happy to walk through it properly, including the parts that went wrong.

Trading bots

Python and Pine Script, part hobby and part laboratory. Hypothesis, backtest, deploy, see what breaks, fix it. Roughly the loop my professional work runs on, with less at stake.

A team should always be able to see the same picture I can.

I am formally trained in EOS, the Entrepreneurial Operating System, and I run it end to end with an annual refresher so everyone knows the playbook. Stripped of the branding it is four habits: a weekly leadership session that solves things in the room, a scorecard everybody can see, honest conversations about whether someone is in the right seat, and big ambitions cut into ninety day pieces.

The reason I care about it is not tidiness. In a growing business the volume of work always outruns the headcount, so the only way quality survives is if the operating model improves faster than the queue does.

My weekly session, drawn to scale

This is the ninety minute global meeting. Strict timing, every item discussed. The point of the shape is that most of it is deliberately protected for solving the biggest problem, not reporting on it.

0–5

Check in. One win, one hard thing.

5–10

Scorecard. Leading indicators only.

10–15

Quarterly goals, pulse check.

15–20

Customer and team headlines.

20–30

Last week's actions. Accountability.

30–85

Identify, Discuss, Solve. Reordered by importance before we start.

85–90

Recap, then rate the meeting out of ten.

Rating the meeting is the part people raise an eyebrow at. It is also the reason the meeting gets better, because a six out of ten in week three is a conversation you would otherwise never have had.

Customers are usually the first to feel a merger

The work is making sure they do not. Four things carry it, and they are in priority order rather than a list of equals.

• People first. The operators carrying the institutional knowledge are the ones you cannot replace, and they are the ones most likely to leave. Psychological safety is not a soft goal here. Honest reporting depends on it, and everything else depends on honest reporting.
• Continuity the customer never notices. Moving legacy systems onto modern, AI supported workflows without a single account feeling the join.
• Protect the accounts most likely to test the door. Uncertainty is when customers shop around, so the largest relationships need more contact, not less.
• Senior people in the difficult rooms. Visibly close to the customer through the transition, especially when the news is not good.

Listen, write it down, then move

The detail changes with every business. The shape rarely does.

Most experienced leaders land somewhere near this eventually. The discipline is holding off on month three until the team can see why.

Six things I believe, and update as I learn.

1. Trust is still the job

   AI changes what it costs to deliver trust. It does not change what trust is made of. The teams I worry about are quietly using it to cut quality and assuming nobody will notice.

2. Resolution beats deflection

   A ticket closed without the underlying problem being solved is not a saving. It moves the cost to the next interaction, and to the customer's patience.

3. The leverage sits between the tiers

   Automating the simplest contacts is largely solved. The interesting work is giving the humans on the harder tiers the same context the bot already had, instantly, so their time goes on the part that genuinely needs a person.

4. Voice of the customer is the underused one

   It is now cheap to read every transcript, ticket and comment and route the patterns back to product. Done properly this is where customer teams earn a seat at the product table.

5. Guardrails first, scale second

   A confidently wrong AI in front of a customer is one of the quickest ways to lose a renewal. Internal channels first, then low stakes external ones, and only then an account that matters.

6. The leadership job is shifting

   Less queue management, more system design. The leaders I learn most from can read a transcript, write a useful prompt, and decide which work should still belong to a person.

The signals I actually trust.

• Curiosity when nobody is watching. Everyone is curious in an interview. I trust it more when someone is still digging into why a number moved on a Friday afternoon.
• Clear writing. Most of this work happens in writing across timezones, and clear writing tends to come from clear thinking. I read someone's written work as carefully as their CV.
• Empathy and commercial discipline in the same person. Most candidates have one. The ones I hire have both and know which to lean on.
• Honest about getting it wrong. I would rather hear about one specific decision someone got wrong, and what they changed, than a polished run of wins.
• Pragmatic about tools. People who pick the right one and put it down again when something better arrives.

Music, a golf swing I am losing to, and a book a fortnight.

When I am not in a spreadsheet or a terminal I am usually making music. It is the part of my week with nothing to do with work, which is most of the reason I keep doing it. I play football and basketball locally, I follow the Premier League, the NBA and the NFL far too closely, and I recently started golf lessons, where I am finding a strange joy in being genuinely terrible at something new. I speak French fluently.

I read about a book a fortnight, mostly murder mysteries and contained world science fiction. Cold cases, sealed silos, ordinary people working out what the system missed. I did not plan for my reading to rhyme with my work, but it plainly does.

• DONEThe Thursday Murder Club, Richard OsmanFour pensioners outpace the police precisely because nobody takes them seriously. A book about being underestimated and the advantages that come with it.
• DONESlow Horses, Mick HerronThe dumping ground for spies who made career ending mistakes. A study of unconventional management and of what actually happens inside a rigid bureaucracy where reputation is everything.
• DONEWool, Hugh HoweyThe last of humanity in a 144 storey underground silo. Truth, power, and how much a society will trade for stability.
• DONEThe Wedding People, Alison EspachDarkly funny, and the best thing I read last year. One unplanned connection redirecting an entire trajectory.
• DONEThe Coming Wave, Mustafa SuleymanHow you contain a technology this cheap to deploy without ending up with either total surveillance or total loss of control. Worth reading if you will be in the room when these tools get adopted at scale.
• NOWThe Marlow Murder Club, Robert ThorogoodA seventy seven year old crossword setter who refuses to let a crime go when the authorities will not take it seriously.