Google-funded Chatbot Planned Over 18 Months
Archant, one of the UK leading publishing companies, faces the same troubles every publishing business faces. The market is changing. The way individuals consume the news is changing. It is time to adapt and innovate, and quickly.
As a company that has been publishing paper content for over 150 years, Archant sits on an enormous amount of local history. Millions of articles are tucked away in their archives, inaccessible to most. What lessons could we learn from them?
That is exactly what Lorna Willis, executive director of digital, data and insight at Archant thought of when she decided to pitch this crazy idea to Google: let’s build a chatbot that knows all there is to know about those last 150 years of recorded news stories.
Not your usual chatbot planning session
Google funds about 200 projects every year through their Digital News Initiative. For this particular project, Archant needed close to £1million from the fund. In order to apply for such a large sum out of their pot, Google requires everyone involved to plan appropriately.
Six months before we even put the application through, ubisend and Archant worked closely together to map the project. There were many moving parts:
- Dusting off the paper archives
- Finding the right digitisation partner
- Organising the data flow, from paper to digital
- Building the artificial intelligence layer
- Making each data point relevant and conversational
All this considering we were six months away from knowing whether or not the project would even be funded.
Phases and key milestones
Our typical chatbot building approach reflects that of a startup validation process. We tend to strip any project that comes to us down to the most simple form. We plan for a minimum viable product (MVP) and up to two further development phases. The MVP and phase one/two usually take two to three months.
In this instance, we had to plan 18 months of development and account for complete unknowns.
We decided to go for an MVP, validation, milestone approach for each phase of the project. That way, we compartment each key phase of the 18-month project as its own smaller project. We split the phases of the project using the biggest pieces.
Archant, ubisend, and Google staff sat down to decide which parts of this project are the biggest unknowns. For instance, the first unknown was the quality of the output from the digitisation scanning. We took this as phase one and implemented our typical approach of MVP, validation, solution and considered the phase over when a key milestone was achieved.
At the time of writing this case study, we are nine months into the project. Our phased approach seems to be working wonders. We have successfully achieved two milestones in the timeframes we planned.
It’s been great to work with ubisend, it is clear that they are a driven and determined team of professionals who provided us with the top quality solution we needed where initially the end product was far from clear.
ubisend’s fresh and enthusiastic approach to the task left us in no doubt that we had picked the idea. strategic partners to work with.
- Chris Amos, Archant.
The Archant’s project is an interesting balance between ‘move fast and break things’ and long-term, sensible planning. Archant took an innovative step into new technology. As technology experts, it is ubisend’s role to make sure this innovative step is as thoughtful as it is audacious.