Andon Group
Book 30 minutes

The people closest to your systems already know what’s wrong.

They knew months ago. The question is what happens when they say so.

Andon is a small group of leaders in their space – engineers, designers, and product people. We come in to figure out what’s actually broken; and when you want us to, we build the fix.

Book 30 minutes
10M+ concurrent users, scaled from 100K
30 engineers led through an acquisition and the integration that followed
Founder & CTO AI product shipped to production, designed and built end to end
Diligence run on acquisition targets
Why Andon

On a Toyota line, any worker can pull the andon cord and stop production when they spot a defect. The cord isn’t the point. The decision behind it is – that letting the person closest to the work stop a line costs less than letting the defect move downstream.

Most engineering orgs have the defects. Not many have the cord.

Some of that is architecture. A system can hide a problem for months, until it’s expensive to fix. The rest is process, which usually comes down to someone who saw it and decided it wasn’t worth saying. It’s almost always both, which is why we don’t treat them as separate practices.

What we do

Two modes: figuring out what’s wrong, and building the solution. Plenty of clients use both.

1Figure out what’s wrong

Architecture review A few weeks in your system, then a plan you can staff.

We spend a few weeks reading the system and talking to the people who maintain it, then a written assessment and a sequenced plan you can staff. You get decision records that withstand the test of time, not a deck that just looks shiny now.

Technical due diligence What you’re actually buying, before you sign.

If you’re buying a company or investing in one, we’ll tell you what you’re actually getting: the code, the architecture, the team, and the distance between the roadmap they showed you and what the system can support. We’ve run diligence on acquisition targets from inside an acquirer, and have also been the company under the microscope. Both views are useful, and most diligence only has one.

AI vendors and lock-in Who owns what the agent learns about your business.

Everyone is buying agentic AI right now, and most of those contracts are silent on who owns what the agent learns about your business. Your procedures, your edge cases, every correction your team made – that accumulates in a vendor’s managed memory, and it usually can’t leave. We’ll tell you what you’re actually being sold, what switching costs eighteen months from now, and what to get in writing before you sign.

Migrations and platform strategy The hard part is data ownership, not services.

Monoliths, shared databases, the rewrite that’s a year past its date – pulling services apart is rarely the hard part, data ownership is. Two teams writing the same tables stay coupled no matter how clean the boundaries look on a diagram, so the ownership map needs to come first or the migration will never finish.

Delivery and process Why three weeks of work takes eleven months.

Why does it take eleven months to ship three weeks of work? The answer is handoffs, unclear ownership, and a feedback loop long enough that nobody connects a decision to its consequence. Rarely is it due to lack of effort.

2Build it

Engineering and design Backend, frontend, UX, product. All of it or one piece.

Sometimes the assessment isn’t the hard part. You know roughly what has to happen; you just don’t have the people. We’ve got backend, frontend, UX, and product, and we’ll take on the whole build or just the parts you prefer – the spike nobody has time for, the integration layer other teams can build against without owning its internals, the UX work that keeps getting pushed, an agent that does something useful.

Either way, it comes back to your engineers with the reasoning intact, not as a black box you now have to maintain.

AI strategy and build Where it actually helps, then shipped.

Most AI projects stall in one of two places. Either nobody can say what the thing is actually for, or someone built a demo that won’t survive production. We’ll work out where AI earns its place in your product and where it doesn’t, then build the part that does: agents, assistants, and the AI features your roadmap keeps deferring.

What the system learns about your business stays in systems you own, the same as everything else we build.

How we work

We don’t take over. Your engineers own what we touch because they’ll be there after we’re gone.

We write things down – assessments, decision records, the tradeoffs with the losing options still in them.

That goes for anything we build with AI. What the system learns about your business lives somewhere you control and can export. If it only works while we’re still on the invoice, it isn’t a system – it’s a subscription.

And we’ll tell you when the answer is cheap. We’d rather spend thirty minutes showing you a simpler fix than sell you six months of work.

Who

Five of us, seventy-five years between us. Architecture, backend, frontend, UX, product, all at the level where people stop asking you how and start asking you whether. We’ve worked together long enough to know who to call for what, so jobs get staffed by what is needed instead of who’s available.

Ben runs it. He’s taken on outside work since 2012, alongside a full-time job: enterprise data architecture, cross-system identity, and integration work across a number of companies. Having experienced such a range of teams and problems with his advisory work, he realized it was the more rewarding part of his career, and eventually he made it his full-time practice.

The full-time job was at an ed-tech startup that McGraw Hill acquired. He stayed through the integration and the decade that followed, ultimately leading a thirty-person engineering org. That’s where the platform went from a hundred thousand concurrent users to over ten million, and where the diligence work started.

Before any of that was two summers at Amazon. During the first one, he noticed that buying a song took more clicks than it needed to, made the case for cutting them, and the change shipped. Conversion moved enough that people paid attention. They brought him back the next summer and offered him a job, which he turned down to go work at a startup instead. An intern catching what a whole team had walked past is more or less the idea behind our name – Andon.

Los Angeles, working with teams anywhere.

Start

Pull the cord.

Thirty minutes, no charge. Tell us what’s holding you back. If we can help, we’ll tell you how. If we can’t, we’ll point you towards what can.

Book 30 minutes
ben@andon.group