Home Services What Is an FDO? How We Choose the Solution Newsletter Podcast Toolkit Press About Contact
How we choose the solution

Existing software, automation, or a custom build.

Most firms recommend the thing they happen to sell. This is the order we actually work in, and what each rung costs you after we leave.

Ysi
Ysi Gonzalez
Founder & Principal Forward Deployed Operator · 8 min read
TL;DR
  • Our work starts a step before the build. The first question is what you're trying to fix — not what we'd like to build. If something already exists that does the job, we're not going to reinvent it.
  • Four routes, in order: change the workflow, buy what already exists, automate across what you own, build custom. Every rung down costs more to own, not just more to build.
  • Before AI is on the table, one question settles it: is your data allowed to be there? In larger companies it's policy, not capability, that decides most AI plans.

We don't sell what we happen to build

Most firms have one answer and go looking for the question. A software house builds software. An integrator installs the platform it resells. A consultancy sells the next phase of the consultancy. None of them are lying to you. They're just answering with the only thing they have.

We're a forward deployment operations firm, so our work starts a step before the build. The opening question isn't what should we build for you — it's what are you trying to fix. From there we plug in whatever the operation actually needs, and bring in the people that particular solution requires, engineers very much included when building turns out to be the answer. Sometimes the right answer is us. Often it isn't, and saying so is the job.

If a product already exists that connects the calendars, we're not going to invent one. Charging you to rebuild something you could subscribe to tomorrow isn't craftsmanship. It's a business model.

The starting question

What are you trying to fix? Not what would we like to build.

The ladder

Once a process is mapped, every step in it comes out at one of four answers. We work down the list in order and stop at the first rung that holds. Going further down is more capable, and more expensive to live with.

Rung 01Change the workflow. No software at all.

Reorder the steps. Remove the approval that exists because of a mistake somebody made three years ago. Move a handoff to the person who already has the information. Nothing to buy, nothing to configure, nothing to roll out.

When it wins

More often than anyone expects. Most steps come back here.

What it costs to own

Nothing. Which is exactly the point.

Rung 02Software that already exists.

Cloud software, chosen and configured around your process. Someone has already built this and sells it to thousands of operations shaped like yours. We implement it, get the real process running inside it, and leave it running.

When it wins

Whenever a real product covers the job. This is the first place we look before anything gets built.

What it costs to own

Implementation, then a subscription. The vendor carries uptime, support, security and the roadmap. When something breaks, there's a company whose job it is to fix it — and it isn't you, and it isn't us. That's why this is the best outcome for everyone involved.

Rung 03Automation and agents across what you already own.

A customized workflow wired across the systems you're already paying for, so records move between them without anyone retyping. Steps that qualify get handed to an agent. Less expensive than a build, and far faster to stand up.

When it wins

The products exist but nothing joins them — or a step is genuinely repetitive, yet too specific to your operation for an off-the-shelf feature to cover.

What it costs to own

Subscriptions for the automation layer, and someone who understands the wiring when a system on either end changes. Smaller than a build. Not zero. A step only gets automated if it's rules-based, runs at least weekly, is low-stakes when it's wrong, and its data already lives in a system — the filter is here. Everything else stays human.

Rung 04A custom build.

Only when nothing above fits: a process genuinely unlike how anyone else works, intellectual property that can't leave your walls, or a data policy that rules the other three out.

When it wins

Rarely — and it should be rare. When it's the right answer it's very much the right answer, and we build it.

What it costs to own

This is the rung people price wrong, so it gets its own section below.

The cost that shows up after launch

The build number is the visible one. The ownership number is the one that should decide.

Custom software has to live somewhere and be looked after by someone, permanently. Either the firm that built it hosts and maintains it — a recurring fee, every year, for as long as you use the thing — or it runs on your own servers, which means you need the internal capacity to keep it standing. If you don't have an IT function, that isn't a line item you can trim. It's a department you'd be starting.

There's a second cost nobody quotes: a custom tool has no vendor roadmap. Nobody is going to ship you a new capability because the market moved. You are the roadmap, and every change is a project with a scope and a price.

None of that makes a custom build wrong. It makes it a decision with a tail — and the tail belongs in the conversation before the build starts, not eighteen months after it ships.

Save this

The build number is the visible one. The ownership number is the one that decides.

Before AI is on the table: is your data allowed to be there?

There's a question that settles more AI projects than any technical constraint, and it almost never gets asked first. Not can AI do this — it usually can. Whether your company permits it to.

An organization that has invested in an enterprise system almost certainly has policies governing what may reach that data. Those policies were written for good reasons and they don't bend because a project is exciting. Public AI platforms learn from what they're given, and you can't control that after the fact. Handing one read-and-write access to a system of record is a governance decision, not a technical one, and it belongs with the people who own that decision — before anyone scopes a build.

Which is why the answer is sometimes no AI at all. A reporting tool that connects straight to the database has no intermediary, refreshes on the system's own schedule, and raises no question about where the data went. If that solves it, that solves it.

And there's a middle route that works more often than people expect. If someone already exports a report to a folder as part of their week, automation can take it from there — pick it up, transform it, and put it on a dashboard that's current every morning. A person still runs the export. It's inexpensive, it holds, and it asks nobody to change a policy.

Occasionally the constraint is itself the reason to build: a tool inside your own environment, with access scoped to exactly what it needs and nothing more. That's a legitimate route to rung four, and one of the few that justifies it outright.

What we ask before proposing anything

A proposal written before these are answered is a guess with a price on it.

Where our own system fits

We do build and license a system of our own, the OpsHub, for CPG operators. It sits on rung four and it exists for a specific case: an off-the-shelf ERP is a real tool, but it's also a nine to eighteen month implementation that somebody has to run while running the operation that needs fixing. Where there's no capacity for that project, a configured system deployed in weeks is the better trade.

It is not the default answer, and we'll tell you when it isn't the answer for you. A firm that only builds will always find a reason to build.

Ask any firm what they'd tell you to buy instead of hiring them. The answer tells you what kind of firm it is.

Ysi (pronounced Easy) is the founder and Principal Forward Deployed Operator at Vantelira, a Montréal-based Forward Deployment Operations firm. Process before tools. Built to last.

Not sure which rung you're on?

The Truth Audit is the entry point. Fourteen days, fixed price, refundable if we don't surface three money leaks. The solution question comes after the process is mapped — never before.

Start Here