What the Week Tells Banks: Three Mistakes to Avoid
What does this week’s news really show?
The week sends a fairly clear signal: banks and large networks are moving towards more AI agents, but without fanfare or slogans. Lelezard highlights CIBC for its AI uses serving the customer experience, Agence Europe reprises the Eurobarometer on rising adoption in EU companies, and Industry Analysts, Inc. pushes the same idea on tools and operations. So we see a deep movement, not an abstract promise.
What matters is the direction taken. AI is no longer confined to committee demonstrations; it enters exchanges, assistance, and request follow-up. In a bank branch, this often means fewer generic answers and more help on concrete cases, at the counter, on the phone, or online.
Still, we must keep our feet on the ground. Not all projects are equal, and a bank can quickly pile up tools without changing the customer journey. But the market clearly pushes towards agents capable of doing more than just sorting tickets or summarizing verbatims.
The first mistake is automating before identifying pain points
In a neighborhood branch in Rabat at noon, the queue moves slowly and the advisor switches screens three times. That’s when you see if automation helps or adds noise. In banking as in insurance, AI fixes nothing if waiting times, call transfers, or advisor clarity are not already visible.
Sometimes there’s a rush to plug in an assistant, a script, a routing, then move quickly on. But if the client waits at the counter, if the contact center bounces calls from one desk to another, or if the explanation remains unclear on a contract, the machine only hides the problem. It can even accelerate it.
The right reflex is to look at friction points as they are experienced daily, not as they appear in a spreadsheet. Waiting, transfer, clarity: these three words often come up when talking with a teller, advisor, or branch manager. As long as they aren’t clearly seen, automation is blind.
I believe we must start from journeys, not tools
My point is clear: the market values teams that manage a complete journey, not those who accumulate layers of automation. CIBC, highlighted this week by The Digital Banker, sends a clear signal on AI use serving the customer experience. Yet this signal must not make us forget the basics: if the journey is poorly managed, automation only puts a veneer on the problem.
This is often seen in banks when a file moves from the website to the branch, then to the contact center. If statuses aren’t clean, reasons aren’t well classified, and follow-up between teams is unclear, you end up with patchwork. Automating before identifying pain points means rushing onto unready ground.
I therefore prefer a very down-to-earth approach: map the journey, identify breaks, then choose the tool that serves that path. It’s slower at first, yes. But this is precisely how you avoid repeated fixes and scripts that don’t hold up to the next rule change.
The limit is here too: a good journey isn’t enough if human execution doesn’t follow. AI helps when it strengthens a clean base, not when it masks process disorder. This discipline makes the difference, not the race for the newest tool.
The three mistakes to avoid this week
- Launching an AI agent without a clear use case. We still see projects start with the tool, then look for the mission afterwards. In a bank, this results in an assistant plugged everywhere but unable to handle a specific file at the counter, call center, or on WhatsApp.
- Measuring only speed. A shorter delay doesn’t say if the client understood the answer, if the advisor properly took over the file, or if the handoff between channels went smoothly. Speed matters, but it’s not enough.
- Forgetting compliance and human channels. If the system doesn’t keep a clean record, doesn’t offer a clear relay to a person, and doesn’t respect local rules, it tires out teams instead of helping them. For us, the sovereign issue isn’t theoretical; Industrie du Maroc often reminds us of this in discussions about dependence on foreign components.
How to act in practice starting next days?
- [object Object]
How to know if AI really helps service?
| Simple measure | What we look at | What it tells the field |
|---|---|---|
| Response time | Time between request and first reply | If AI really speeds up, or adds one more step |
| Customer callback | Having to recontact to move forward | If the file stays fluid, or returns to the start |
| Resolution at first contact | The problem solved from the first interaction | If the tool helps decide, not just respond |
| Reopenings | Cases reopened after a too-quick closure | If the response was thorough, or just rushed |
| Repeated complaint | The same irritants recurring in verbatim and on the phone | If the AI addresses the cause, not just the symptom |
| Team workload | What advisors, managers, and back offices experience | If the AI reduces noise, or creates it elsewhere |
These measures are enough to read the week without getting lost in overloaded tables. They can be tracked by agency, channel, and request type, with JABB VOC if needed. The real test remains simple: fewer unnecessary returns, fewer lingering cases, and teams that breathe a little easier.
If you want to connect the voice of the customer, alerts, and service journeys in one tracking, see how JABB structures these signals.
