Turning customer signal into systems: 2.5 years at Amazon

Two and a half years ago, I took a role I'd never tried before: something on the external-facing, customer side. Up until then, I'd always been on the backend — building the product, researching the customer in the data (but not talking to them directly). I'd had a period as a founder where I did every role, but even then I was primarily on the R&D and partnerships side, rather than directly with customers.

So in 2024, when the opportunity presented itself, I decided: you know what, let's talk directly to customers and see what we can learn.

I took a role as a CSM at Amazon, and I remember someone asking whether the role was essentially customer support. It included problem-solving, certainly, but the more interesting responsibility was recognizing which customer problems were isolated and which revealed a system that needed to change.

Really, it was two and a half years of learning to read customer signal — and turn it into something that scales. The listening mattered enormously, but for me it was always in service of the build: understand the problem deeply, then go solve it for good.

My goal as a CSM wasn't to work with one large client the whole time. I wanted variety — as many customers as possible, actually — so I could get a breadth of what each one wanted from Amazon. It doesn't matter the industry or the product they sell. What matters more is: who is the leadership team, how does the org structure function, who is the decision maker (sometimes these are all separate people), what is their growth strategy, what are their top pain points and weaknesses, and are their strengths being leveraged to their full capability? That's what's important — not whether they're selling sponges, beans, computers, or data center shelves.

As a CSM, I sat at the intersection of strategy and execution: translating a customer's goals into adoption, operational problem-solving, and changes the broader system could support.

The most interesting thing to learn was the Amazon side of things. Amazon is a universe of internal structures and systems, layered on top of and next to each other. Navigating the company is challenging and overwhelming for new hires — though if you've been a founder, it doesn't seem quite so hard. In fact, you have a higher chance of getting a reply from another Amazonian than from an investor trying to ghost you. And getting the attention of other teams — already swamped, with no incentive to reply — requires storytelling. Which, if you've fundraised, you already know how to do.

What I found interesting about the CSM role is that, when I started, a lot of the work was still manual. In any big enterprise, knowledge lives in endless SOPs and wikis — and just as often in people's heads, shared peer to peer — so half the job is finding the right person to point you in the right direction. That's exactly the kind of problem AI is now collapsing across the industry: taking scattered information and pulling it into a single, askable place.

Something else I found interesting: many customer questions had documented answers somewhere — in public guidance, internal systems, or Vendor Central — but finding the right answer and understanding how it applied to a specific business still took time. Customers were not lacking intelligence; they were navigating an enormous volume of fragmented information. What they needed was synthesis: a clear answer, grounded in the right source, delivered in the context of their business. But once you've delivered the same handful of answers enough times, the interesting question stops being "what's the answer?" and becomes "why am I still the one answering this — and what would it take to build something that answers it once, for everyone?"

I think that's where AI can add tremendous value. It can take a huge amount of information, summarize it in an easy-to-understand way, and give you the actual information you need. It turns research at the minutiae level into something conversational — and if it's conversational, it's easier to digest.

It also taught me — again — how important storytelling is to customer adoption. As the recurring operational questions became easier to resolve, our meetings became shorter and more focused. That was evidence that the systems were working: less time spent locating information, more time available for decisions that genuinely required the customer and CSM together.

This past weekend, after officially leaving Amazon, I decided it would be fun to build HENRY — Helpful Expert, Navigating Retail Yield — using purely public, online data (no confidential information) as a one-stop shop for vendors' most common questions. It prototypes the repeatable research and diagnostic layer of the CSM role: answering common operational questions, structuring financial analysis, and helping vendors arrive at conversations better informed. If it were ever in full-scale production, the goal is for it to be an easy repository for the same concerns, questions, and comments that so many of my customers had — all of which are solvable.

I liked the people, genuinely — being a CSM is a bit like being a business therapist. Watching a customer grow is its own reward. The relationship mattered because it revealed context no dashboard could provide: what the customer feared, what leadership valued, and what constraints shaped the decision. The best systems preserve that context while removing the repetitive work surrounding it. The satisfying part came from noticing the same question surface again and again — and making sure it never had to be asked twice. Over time the vendors needed me less and less, and that was the point. The goal was not to remove the human relationship. It was to make customers less dependent on a person for repeatable research and routine answers. If technology could return the hours once spent searching SOPs or repeating the same explanation, the customer and CSM could spend that time on judgment, growth, and problems that genuinely required another person.

What I'm really excited about — and think will be impactful — is the future of AI in work. It's all in search of one greater goal: how can we spend less time on things we don't want to be doing, and more time on the things where we can have real impact and use our creativity, instead of grinding through manual process?

That is the work I want to keep doing: staying close enough to understand the customer's problem, then building the system that solves it at scale.