You need a piece of software your business does not have. Maybe it is a customer portal, an internal tool, an integration between two systems, or a new AI workflow. The question is the same in every case. Do you build it in-house, hire an agency to build it for you, or work with a partner who scopes, builds, and stays involved after launch?
There is no universal right answer. The right answer depends on three things: the strategic value of the work, the speed you need, and the team you already have. Get that read wrong and you will waste somewhere between 3 and 9 months figuring out what you should have known on day one.
This is the framework we walk clients through in the first hour of nearly every project we take on.
The three options, honestly
Each of these paths is good for something. Each is bad for something else. The mistake teams make is choosing the path based on cost alone, not based on fit.
In-house
You hire engineers, designers, and product people and build the thing yourself. You own the code, the team, and the institutional knowledge. You also own the hiring, the management, and the retention problem.
This is the right call when the software is a long-term strategic asset, when the team will keep iterating on it for years, and when the work is core to how your business differentiates. Harvard Business Review and other strategy publications have covered the build-versus-buy logic in depth across the last decade.
It is the wrong call when you need it shipped this quarter and your hiring funnel is six months long, or when the work is a one-off project that will not need a permanent team after launch.
Agency
You hire a development shop that takes a spec, builds it, and hands it back. You pay a fixed price (or a time and materials rate), and you get a delivered product. Most agencies do not stay involved after delivery, and most do not push back on your spec.
This works when the spec is genuinely clear, the work is bounded, and you have someone internally who can own the result after handoff. It does not work when the spec is fuzzy, the project keeps evolving, or you are buying expertise rather than execution.
The risk with agencies is that the incentives can pull against the outcome. They are paid to deliver against the spec, not to make sure the spec was right. The Standish Group’s long-running CHAOS research has consistently found that the majority of fixed-scope software projects either go over budget, miss their timelines, or fail outright. The driver is almost always that the original spec was wrong, and nobody was incentivized to catch it.
Partner
A partner is closer to an embedded team. They start with an audit, push back on the spec, scope the work honestly, build it, and stay involved after launch. They are usually smaller than an agency, more senior, and more expensive on a per-hour basis. The total project cost is often similar to or lower than an agency, because the partner does less rework and ships closer to what you actually needed.
This is the right call for most growing businesses doing AI implementation, custom software, or any project where the requirements will evolve as you learn. It is the wrong call for genuinely commodity work, like a standard ecommerce store on Shopify, where an agency or a Shopify expert is faster and cheaper.
Real cost ranges in 2026
Numbers vary by region, complexity, and how senior the team is. These are reasonable ballparks for SMB and mid-market work in North America.
In-house (loaded cost per engineer). A senior full-stack engineer costs roughly 175,000 to 250,000 dollars per year fully loaded once you include benefits, equipment, and management overhead. A small team of three engineers and a designer runs 750k to 1.2M per year. That is your floor.
Agency. Most reputable agencies in the US bill between 175 and 350 dollars per hour for senior work. A typical 3 month MVP project lands between 80,000 and 250,000 dollars. AI-augmented projects often run higher because of integration complexity.
Partner. Hourly or weekly rates similar to agencies, but the engagements are typically smaller because the scope is sharper. A typical project lands between 30,000 and 150,000 dollars, with an upfront audit phase at a fixed price so you know what you are buying before committing.
These ranges are honest, not optimistic. If a vendor quotes 15,000 dollars for a real custom software build, ask what they are leaving out.
Five questions that pick the answer for you
Walk through these and the right path becomes obvious in most cases.
- Will this software be a long-term differentiator? If yes, lean in-house. If no, lean partner or agency. Differentiating work belongs to the company that benefits from it.
- Do you have an internal owner? If you cannot name a single person who will own the result after launch, do not start. The build will not stick.
- Is the spec clear and stable? Clear and stable means an agency works. Fuzzy or evolving means a partner. A genuinely commodity spec means buying a SaaS product, not building.
- What is your timeline? Hiring an in-house team takes 4 to 8 months before the first commit. If you need shipped software in 90 days, that is not the path.
- How important is the post-launch period? If the software has to keep evolving for two years, you need a team that survives that. That is in-house, or a partner with a real ongoing relationship. An agency that disappears after launch will leave you supporting code you do not understand.
Hidden costs of each path
Every path has costs that do not show up in the proposal.
In-house hidden costs. Hiring (recruiting, comp benchmarking, interview time, signing bonuses). Management overhead. Tooling and infrastructure. The 4 to 8 month ramp before the first useful ship. The retention problem 18 months in. According to Built In’s compensation reports, engineer comp keeps climbing year over year.
Agency hidden costs. Change orders. The handoff cost when the agency leaves and your team has to learn the codebase from scratch. The mismatch between what you actually needed and what was in the spec.
Partner hidden costs. Less price elasticity (smaller teams, less ability to discount). Smaller bench, so timelines depend on team availability. You usually cannot scale to 12 engineers overnight if the project balloons.
When to switch paths mid-project
This is allowed. Switching paths is usually cheaper than finishing on the wrong one.
Switch from agency to partner when the spec has changed three times and the agency keeps writing change orders. The agency is doing what they are paid to do. You needed someone pushing back two months ago.
Switch from in-house to partner when the team you hired is great, but the work has stalled because of process or skills gaps. A partner can unstick the project while your team levels up.
Switch from partner to in-house when the project succeeds and becomes core to the business. The partner has done their job. Now hire the team to own it long term.
What to do next
Pick the right path for the project in front of you. Not the path that feels safe, and not the path your CFO will approve fastest.
If you want to talk through a specific project and which path actually fits, the AI Clarity Audit is built for exactly this situation. It gives you a written, honest read on scope, risk, and the cheapest path to a working result, whether or not we are the ones who build it.
Send a project inquiry and we can walk through where you are stuck.
