Docplanner Phone is an omnichannel platform for Clinics. It allows medical staff to centralise all patient communications (Phone, Whatsapp, SMS, forms), and solve one by one thanks to inbox-zero approach and availability of async channels.
The project involved understanding painpoints of Clinic secretary around most repetitive and painful jobs they have.
PROBLEM SPACE
The problem we wanted to solve with this one, was helping secretaries from the most mundane day-to-day tasks, so that they can increase their productivity and focus on things that matter most to them.
To understand those problems better, I was checking Hotjar recordings of their Inbox behaviours: They were really acting super fast in there, having to multitask all the time, while on the calls with patients.
I also established and used for this purpose Continuous Discovery interviews, where we were asking what are the secretary jobs and which ones they found most repetitive, time-consuming, and frustrating – we also asked about sentiments around the topic of AI, which we wanted to employ on the side of solutions. Spending time on understanding what given conversation was about was often mentioned. And the users were mostly really positive with AI that would actually be the help for them to use.
CALL CATEGORIES
Hypothesis: "Secretaries will be more likely to close conversations with a category if they do not need to choose one of options"
Currently, call categories are a manual work required from secretaries. This can be just one click, but still sometimes it is one click too much – so the result is only approx. 50% of all conversations end with a subject set. Also the number of unique users of "set subject" is not growing as fast as the global number of users.
In this solution, we pre-define the category, and only ask the secretary to Close conversation with such category (or alternatively click on the category to display the list and choose other). Among alternative ideas for the solution was using more flexible Tags system, but the idea would require much more effort to validate the hypothesis.
REPLAYING CALLS
Hypothesis: "Secretaries will be able to navigate and close conversations faster, having an opportunity to read summary instead of listening to full recordings of calls"
Currently, secretaries go back often to recordings often, to understand what the patient wanted when they talked to them in the past or someone else answered the call, and they need to take over. This behaviour is present in almost all of our users.
In the solution, call summaries can be accepted, rejected, or edited, and the reject action triggers Edit modal to appear – as most often this will be the recommended next action.
NAMING AI HELPER
I proposed one setting and universal name for all AI features (then in closed beta) that would be an easy to understand umbrella term, also letting us easily refer to it in the interface, not having to repeat legal consents and so on.
The name I proposed was "Assistant", as a human-like assistant for the secretaries in their day to day, wanting to both stress the help they can get, and not using AI in it, not to scare away the sceptics, if any.
WHAT NEXT?
Right now, we’re getting more and more signals that our clients want AI to help them – and as they already sign up for our Closed beta organically – we’re scaling call summaries to all customers.
Call categorisation is unfortunately more complex to develop and achieve 90% accuracy which was our goal before scaling to clients, so we'll train different AI models and hopefully scale this one as well soon.
Feedback about Call Summaries from closed beta users is super positive and we were observing more and more events such as accept/reject/edit - highest growth being in approve, 574 events in January, compared to half that number in December for the same number of users. Based on this, it has become one of value propositions in the new tiered pricing plans.