Teppo Kuisma // November 19 2018
Why AI/ML Sprints rock?
Most of us have experienced the fast paced and result-oriented experience on running a Design Sprint. From the time of Jake Knapp published the Sprint – How to solve big problems and Test New Ideas in Just Five Days, it’s been always a pleasure to run sprints or be part of them. At Digitalist we’ve run more than 100 Sprints so far. It’s been great to witness the well-structured debate around the methodology. Should a sprint be run on a week or over a week – or can it be just 3 days? If you are planning to Dent / Disrupt / do some Ideas to Life with us, ast Digitalist we do approach things from heavy co-creation and research form. We’ve learned to do Sprints very well. (Hint; be in touch if you want to run a really good one)
Recently many friends, clients and innovators have asked me about the best way to start a Deep Learning or Artificial Intelligence project. Classical Data Scientist Pro answer is of course “Spend the next 6 months of cleaning some essential data, and let’s meet after that”. While in some scenarios that’s what needs to be done, it’s still not the fastest nor most efficient way to start. Yes, Deep Learning and Neural Networks are mind-blowingly powerful and need good quality data. But very few organisations have the discipline / possibility to start cleaning up their data strategies when the end results or even the target is still unknown.
So could we solve the problem by approaching AI challenge’s from a very lean Design Sprint perspective? Could the target of getting a clickable prototype up to end users be so valuable that it would justify even some cheating (smaller sample size, less variables) on the actual data?
It’s important to understand that in every full-blown Design Sprint we’ve experienced, people actually discover together majorly on three areas;
- The Methodology of planning together in accelerated, lean mode. A.k.a Design Sprint
- The idea, change, solution itself
- The team taking part in the planning.
In the context of using the Design Sprint to drive the AI driven possibilities forward in organization, the setup is slightly different. There’s an inherent assumption that a new technological development model (AI, Deep Learning) will open new opportunities. This means there’s a fourth (4.) strong driver for the activity – Getting to know the possibilities of AI.
This far, we’ve had only positive experiences. Eye-opening moments as our cocreative teams have worked with our clients. Late evenings with hard thinking on “could this really be done?”. Long nights creating prototypes. The adrenaline rush of getting to show it to users. Sprints really work and we’ve learned a lot.
Like with all the development work, it’s important to be careful with the weapon choice. Not all things can be prototyped in 24-36 hours. Not all problems are solvable in a week.
But what comes to From Ideas to Life, experimenting with something new – definitely worth the week. Like one of our participants told “Typically this amount of decision making and shared organizational learning would have taken us 6-8months, now it took two weeks in calendar time!”
Quite an efficient way, isn’t it?
If you find yourself interested, don’t hesitate to contact us!
Written while listening the “You are the voice” by John Farnham.