UX agency or in-house team: how to keep the skill after the project

On-Off Group

September 2026

Hire an agency when you need the work shipped now, and build the in-house team's skill at the same time by having them work on it with you. If the agency leaves and nobody on your side can run the next round, you bought output, not skill.

If you are weighing up a UX agency or an in-house team for a customer journey redesign, this is for you. It is for anyone who has to show results this month and has already paid for a project that ended in a slide pack.

The choice usually gets framed two ways. Hire an agency and get screens back. Hire or borrow designers and wait for them to get up to speed. There is a third option: run the project with your own people in every session, so they can do the next customer journey without buying it again.

Here is how that runs, and what to check before you sign anything.

What it looks like when the agency leaves

The redesign lands. New onboarding flow, new screens, a Figma file and a handover call. Three months later someone in branch operations asks for a small change to the requirements step. Nobody in your team knows why the field order was chosen, so the change gets argued about for weeks and then goes into a backlog.

Six months later there is a second journey to fix. The knowledge of how the first one was researched, tested and cut down is sitting in an agency that has moved on to another client.

That cost rarely shows up in the quote. You paid for a redesign and you got one. You did not buy the ability to do the next one. That is fine for a one-off page you will never touch again. It is expensive when you have a list of customer journeys to fix and a date on all of them.

Decide what you want to own before you write the brief

Before you compare suppliers, write down what your team should be able to do without help in twelve months. Things like: recruit five real customers and watch them use the flow on their own. Run a session that turns messy feedback into a list in priority order. Write a prompt that gets a usable first draft of copy or test questions. And judge when a rough version is good enough to put in front of someone, which is the part most people get stuck on.

That list changes what you buy. An agency brief that says "redesign the onboarding journey" gets you screens. A brief that says "redesign the onboarding journey with two of my people in the room for every session, and leave them able to run the next one" gets you screens and two people who can repeat it.

It also changes who you shortlist. Ask each firm to describe the last time a client team ran the next round on their own. You find out fast who has actually done it.

Your people do the work, with help sitting next to them

The way people pick this up is not complicated. Your people do the work, with someone experienced next to them, on a real project with a real deadline.

So the researcher writes the screener and the agency reviews it, rather than the other way round. Your product owner moderates two of the five user sessions. Your developer builds the rough version, not a polished prototype made elsewhere. The first go will be slower and messier than if the supplier just did it. That slower first round costs you time once, not every time.

We have run workshops for Philippine banks and insurers since 2015. People who sat in a Sun Life workshop that first year are now senior across the industry. None of them mention the handover documents. They remember running a session badly and then running it again properly.

Get a before number, whoever does the work

The problem is usually not the design work. It is that there is no before number, so when your boss asks if it worked, all you can do is shrug.

Pick one number before anyone touches a screen. Completion rate on the onboarding flow. Turnaround time from submission to approved. Calls to the branch about one step. Measure it for two weeks as it is today, with the current flow, and write it down somewhere your boss can see.

That number protects you both ways. If the redesign moves it, you have proof that did not come from a supplier's slide. If it barely moves, you know in a few weeks instead of defending the project for a year. Small honest numbers hold up in front of management. Big vague ones get picked apart.

And your team keeps it. Getting a before number costs nothing except remembering to do it, and once someone has done it they tend not to skip it again.

Short projects teach more than long ones

On a long project the feedback turns up months after the decision, when nobody remembers why they chose it.

A large insurer had customer-facing web journeys that took six months or more to get designed and live. Running them as five day design sprints, the journeys were designed and tested with customers in two weeks and live in four. Time to market went from 26 weeks to 4. Completion rates on the new journeys rose 80 per cent, and the drop-off points that had been there before were gone. The client measured it, not us.

The numbers are not the only thing that came out of it. In four weeks the team went all the way round research, build, test and fix, and saw which of their assumptions customers ignored. That is what changes how someone works on Monday. A classroom version does not do it.

There is a write-up of a similar sprint with Meralco if you want to see how the week is structured: design sprint with Meralco.

Training on its own has the same problem in reverse

The in-house route usually starts with a course. Everyone attends, the feedback scores are good, and nothing changes in the next sprint. Training that people are made to attend does not change how they work on Monday.

Attach the training to a live piece of work that has a date on it. Teach it on the customer journey you have already promised to fix. Then the first real task people do after the training is the one they were trained on.

What to do this week

Write down what you want your own team able to do without help in twelve months.
Pick one number on the journey you are fixing next and start measuring it now, before any design work.
Rewrite your supplier brief so your people are named in the sessions, not just cc'd on the output.
Ask any shortlisted firm for an example where the client team ran the next round themselves.

Book the first user session now, even if the flow is not ready.

More of the projects behind this, including the sprint write-ups, are at onoffgroup.com.

Questions people ask

How do I know if the skill actually stayed with my team?
Look at the first real piece of work after the agency leaves. Did someone on your team run the next round of testing or the next design change without calling for help?

Why not just send people on training instead?
Training people are made to attend does not change how they work on Monday. Skill sticks when the team does the work on a live project with someone alongside them.

What should I agree with the agency up front?
Agree who from your team sits in the work, not just the readouts, and what they will be able to run on their own by the end.

How do I show my boss it worked?
Take a number before you start, such as turnaround time on the current flow, and measure the same thing after. Without a before number, the answer to did it work is a shrug.

On-Off Group trains teams, tests products with real customers, finds where a transformation has stalled and builds what gets it moving, for banks, insurers and enterprises in the Philippines, since 2015. How we help with ux training.