Your partner's work does not end at go-live. It changes. A good partner trains your team on the new way of working, watches how people actually use it and fixes the exceptions it cannot handle. They measure results against the numbers you had before launch and help you pick the next process. If the project ended with a login and a goodbye, you bought software, not a working process.
Who this is for: owners, COOs and operations leaders of established companies who are choosing an AI implementation partner, or whose project launched a few weeks ago and usage is dropping. The examples use an accounting firm that automated how client tax documents come in. The same steps work for an insurance office or any professional services firm.
How we know: the report quoted below, and the process and project work our founder, Alvin Chikaeze, did before FlowChainLabs inside Southern Glazer's Wine & Spirits, PepsiCo, Stanley 1913, City Electric Supply, AT&T and Salesforce.
Why do AI projects lose steam after launch?
One report found that of the challenges organizations face in implementing AI initiatives, only ten percent involved the AI algorithms, and twenty percent came from technology problems. The same report found that seventy-four percent of organizations have yet to show tangible value from their use of AI.
So plan for the people and the process, not only the software. That work does not stop at launch. It starts there, because launch is the first time the system meets your real work.
Take the accounting firm. Before launch, the new intake flow was tested on clean files. In the first week of tax season it meets a mortgage statement scanned at an angle and a brokerage statement locked with a password. It meets a client who emails documents instead of uploading them and a spouse who sends the same file three times. Each one needs a rule or a person. If nobody owns them, your staff quietly go back to the old way, and you pay for two processes instead of one.
What should the handoff include?
A handoff is not a login and a training video. You should be able to point to each of these.
- The written process. Step by step: what the automation does, what a person does and where one hands off to the other. Written for the person doing the work, not for the builder.
- An owner for every exception. A list of the cases the system cannot handle and the person who handles each one. At the accounting firm, an unreadable scan goes to the intake coordinator and a document for the wrong tax year goes to the preparer. A client who has not uploaded anything after two reminders goes to the partner on the account.
- How to change a rule. Who is allowed to change it, how they test the change before it touches client files and where the change is recorded.
- What happens when it breaks. Who gets told, how they find out and what staff do in the meantime.
- Training by role. Intake staff, preparers and partners each touch it differently, so each role gets its own short guide.
If any of these is missing, the gap lands on your best employee. That is how a new system ends up living in one person's head.
How can you tell whether your team is actually using it?
Ask three questions. You do not need a dashboard to answer them.
Is the work going through the new way?
Count how much work went through the new flow last week and how much still came in the old way. At the accounting firm: how many client documents came through the upload flow, and how many still arrived by email or on paper? If the old channel is growing, usage is slipping.
What is happening to the exceptions?
Open the exception queue. Look at how many items are in it, how old the oldest one is and who is clearing it. A healthy queue shrinks as the rules improve. A queue that keeps growing, or that nobody owns, means the system is quietly pushing work back onto people.
Is the old workaround coming back?
This is the clearest sign. The shared inbox people still check just in case. The spreadsheet tracker someone kept. The sticky notes on a monitor. Ask three people to show you how they did the task this morning. If they open the old tool first, they do not trust the new one yet. Ask why: when a real exception is not handled, the workaround is how the work still gets done.
When it works, the old channel shrinks, the exception queue stays small and gets cleared every day, and nobody keeps a private copy of the work.
What should you measure before and after go-live?
You cannot tell whether a project worked without a before number. Take it before launch. If nobody did, take it now from last season's records, email history or time sheets.
| What to measure | Where the before number comes from | What you want to see after |
|---|---|---|
| Time from a client's first document to a file ready for a preparer | Dates in your practice software or inbox | Shorter, and steady through the busy weeks |
| Staff hours spent chasing missing documents each week | Ask the intake team to log a normal week | Fewer hours, moved to work that needs a person |
| Files sent back because something was missing | Preparer notes or review comments | Fewer returns |
| Share of documents that come through the new flow | Count by channel for one week | Rising, then holding |
| Exceptions each week and how long each one waits | The exception queue | Fewer over time, none left waiting for days |
Measure the outcome, not the activity. A count of documents processed proves the system ran. Shorter turnaround and fewer hours chasing clients prove it paid off. Our guide on how to measure AI in your business walks through setting these up.
What should the partner keep doing after launch?
- Train again after real use. The first training happens before launch. The second comes a few weeks in, built on what people actually did and the questions they actually asked. New hires need it too. See how we approach team training.
- Watch how people work. Sit with the people who use it, read the exception queue and look at where work leaks back to the old way.
- Fix exceptions at the root. When the same exception keeps showing up, it becomes a new rule, a clearer instruction or a named owner. You should hear what changed and why.
- Report against the baseline. A short monthly report on the measures above, with the before number next to the after number. If a number got worse, the report says so.
- Help pick the next process. Once the first one holds, the partner helps you choose what to fix next, using what the first project taught you about your data and your team.
What are the warning signs of a partner who hands over a login and leaves?
- Support means a ticket form, and no named person knows your setup.
- Nobody asked for your before numbers, so nobody can show a result.
- Only the partner can change a rule, and every change takes weeks.
- Training was one recorded session on launch day.
- The monthly update counts activity, like tasks run, and never mentions exceptions or the old workaround.
- They pitch the next project before anyone is using the first one.
Which questions should you ask a partner before you sign?
| Ask this | A good answer | A warning sign |
|---|---|---|
| Who do we call when it handles a case wrong? | A named person who knows our setup, and how fast they respond | A help center link |
| What does the handoff include? | The written process, an owner for each exception and how to change a rule | A login and a video |
| How will we know people are using it? | They track usage, exceptions and the old channel | They say adoption is up to us |
| What baseline will you take before launch? | Named measures, taken from our records before go-live | They only measure after launch |
| How do we change a rule ourselves? | A clear process with a test step | Only their team can touch it |
| What happens in the months after launch? | Follow-up training, regular reviews of real use and a monthly report | Nothing is planned after launch |
What can you check this week?
- Pull the last two weeks of work the new system was built for. Count how much went through the new flow and how much came in the old way.
- Open the exception queue. Note the oldest item and who owns it.
- Ask three people to show you how they did the task this morning. Watch for the old inbox, spreadsheet or notes.
- Find your before numbers. If you have none, write down this week's numbers now so next month has something to compare against.
- Ask your partner one question: if we need to change a rule today, who does it and how long does it take?
If the new flow carries most of the work, the queue is owned and a rule can change quickly, keep going. If not, you know where the gap is.
How does FlowChainLabs handle what happens after go-live?
We map how your business runs and build AI into it. We build it into the systems you already run, teach your team to use it and keep improving it. See how that works on our AI implementation page.
Watching usage drop on something you launched, or about to choose a partner? Schedule a strategy call. Bring your process notes and any numbers you have from before launch.
