Small AI features beyond chatbots: four myths about customer journeys
Small AI features beyond chatbots sit inside a step people already use, like reading an uploaded ID to fill in a form. Pick one step people struggle with, time it first, and fix the journey before adding AI.
This is for anyone planning a first AI feature in a customer journey, whether it is for customers on the app or for branch staff behind the counter. If you have been searching for small AI features beyond chatbots, you are already asking a better question than most planning meetings do.
The same four beliefs come up whenever a client team plans a first AI feature. Each one sounds sensible in the room, and each one slows the work down or points it at the wrong step.
Below is each belief, what it looks like on a normal working day, and what we would do instead, with one example for each. There is a short list at the end for this week.
Myth 1: adding AI means putting a chatbot on the website
Someone in a meeting says the customer journey needs AI, and a few minutes later the plan is a chat window in the corner of the website.
Picture a customer halfway through an onboarding form. They get stuck on a field. To get help, they open the chat, type out what is wrong, wait for a reply, then go back to the form and try to find their place. Plenty of them never make it back.
A better place for AI is inside the step they are already on. A customer uploads a photo of their ID. The AI reads the name, birthday and address and fills in the form fields. The customer checks the details and taps next. No chat window, no explaining themselves.
The same idea works for branch staff. If staff read details off a document and type them into a screen, AI can do the reading and leave the checking to them.
These are small AI features beyond chatbots, and they are easier to test, because you can see whether the step got faster.
Myth 2: a first AI feature has to be big to be worth it
The first AI idea often arrives as a platform. There is a roadmap, a vendor shortlist and a timeline measured in months. Plenty of people have sat through a six-month project like that which ended in a slide pack and nothing anyone could measure.
One small step is enough to start. Look for the step where customers give up, or where branch staff type the same details twice because two systems do not talk to each other.
Then write down the turnaround time for that step before you build anything. How long does it take today, from the customer starting the form to the account being opened? How many minutes does a staff member spend on each application? Note how you measured it, so you can measure it the same way afterwards.
Without that number, when your boss asks if the feature worked, the honest answer is a shrug. With it, even a small change is something you can put in front of them. If the improvement turns out small, say so. A small honest number is easier to defend than a big vague claim.
Myth 3: we can't build anything until the business case is signed off
A lot of business cases for AI features are written from guesses. Estimated time saved, estimated uptake, a projected figure on the third page of a PowerPoint. The approvers question the guesses, the document goes round again, and nothing gets built while it does.
Build a rough working version of the feature first. It does not need to connect to your core systems. It needs to do the one thing well enough for a customer to try it, such as reading an uploaded ID and filling in three fields. A rough version like that can often be put together in an evening.
Then watch a few customers use it. We usually watch five real people. Note how many finish, where they hesitate, and how long the step takes compared with your before number.
Now the business case can point to what people actually did with the feature. Approvers argue less with a video of a customer finishing a form than with an estimate.
Myth 4: AI will smooth over the clunky parts of our journey
This one usually starts with a complaint. The onboarding form asks for the same document twice. Customers get annoyed, some leave, and branch staff hear about it every day. The proposed answer is a chatbot that explains why the form asks twice.
The chatbot does not fix the form. It gives the failure a friendlier voice. AI added to a broken customer journey makes the same failure happen faster.
Fix the journey first, then decide which step still needs AI.
A large insurer we worked with ran a series of five-day design sprints on customer-facing web journeys. Changes that used to take six months or more were designed and tested with customers in two weeks and were live in four. Completion rates on the new journeys rose 80 percent, and the previous drop-off points disappeared. The client measured those numbers themselves.
None of that needed AI. Once a journey works, the step that still needs AI is easier to spot, and usually smaller than anyone expected.
What to do this week
- Pick one step in one customer journey where customers give up or branch staff retype the same details.
- Time that step as it runs today, and write down the turnaround time and how you measured it.
- Look for anything in the step that is simply broken, like a document asked for twice, and fix that before adding AI.
- Build a rough version of one AI feature inside that step, such as reading an uploaded ID to fill in the form.
- Watch five people use it, and note how many finish and where they stop.
Questions people ask
What is an example of a small AI feature that is not a chatbot?
Reading a photo of an uploaded ID and filling in the name, birthday and address on the form. The customer checks the details and carries on, with no chat window involved.
How small should a first AI feature be?
One step in one customer journey is enough. Choose the step where customers give up or where branch staff type the same details twice.
Why measure turnaround time before building anything?
Without a before number, nobody can show whether the feature changed anything. With it, even a small improvement is something you can show your boss.
Should we add AI or fix the customer journey first?
Fix the journey first. AI added to a broken journey makes the same failure happen faster, and once the journey works, the step that still needs AI is easier to see.
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. Who we are.
