We need a chatbot! What Now?
Every ten years, we seem to leverage new technology driven curiosity. We iterate through discover, pilot, learn. Eventually we end up in sustainable business impact, Andreas Sjöström VP Digital Advisory Services, considers how best to get started with your Chatbot.
1998: “We need a homepage!”
2008: “We need an app!”
2018: “We need a chatbot!”
Most know it’s inevitable. Few really know why. Fewer know how.
Every ten years, we seem to leverage new technology driven curiosity. We iterate through discover, pilot, learn. Eventually we end up in sustainable business impact.
My advice is to not oversell the first few projects. Believe in them and enjoy the ride of piloting, but don’t overspend. Don’t kill the bot baby, instead evolve it. Try with real users, real customers, real people.
As with the first homepages (1998) and apps (2008), some won’t see any use and some will even mock. Don’t be the latter. Focus on usefulness and push on.
Some bots today are obviously just simple IF-statements, just as some of the first homepages just were static brochures. Other bots are just a simplified menu system, exactly as some of the first apps. Eventually and if you sincerely focus on making this happen, your bot will deliver on the true cognitive promises of being:
- really understanding both what you say and mean, in context
- really learning from its conversations
- really being able to connect to unstructured and unknown data sources on its own
My advice for the initial few phases:
- Start with a narrow but business relevant scope
- Identify what is new and what is not; for example natural language processing is new (compared with apps), context relevance is not. Learn and/recruit for the new.
- Involve real users / customers
- Understand the importance of architecture
- Master data management
- Omnichannel enabled digital platform, APIs
- Don’t be afraid to experiment with multiple AI/ML/cognitive platforms at the same time
- IBM Watson, Microsoft Cortana and Azure Bot Service, Google Cloud AI and Prediction, Dialogflow, Wit.ai, Amazon Lex, to name a few leading alternatives
… and most importantly, have fun. Life’s too short to not enjoy the ride.