Most of what we read about AI success relates to large digital-native companies like Google and Facebook -- businesses built in the digital era around using data to target massive consumer audiences.
But what about the digital transformers? The world-leading businesses born in the pre-digital age that are now undergoing digital transformation.
These companies underpin the world’s manufacturing, engineering, energy and transportation services. They depend upon R&D, operational processes and physical assets. They are grounded in the physical world, with products that are tangible and finite.
Most now share a determination to digitalize their businesses in order to push into new, high-value marketplaces, increase differentiation and avoid the ever-increasing threat of disruption. They are asking: "How do we create value and competitive advantage from digitalization and AI?"
AI As Part Of A Digital Transformation
Digitalization is key to taking advantage of AI, but it does not automatically make you AI-ready. Companies need to digitalize in a way that sets them up to take advantage of AI.
The data these companies collect through digitalization comes largely from R&D and industrial processes, not large swaths of users. The type of business insight this data can provide, and the way data needs to be handled, is completely different from the consumer insights that companies like Google, Facebook and Amazon profit from.
There are lessons industry stalwarts can learn from digital-native companies. But they are starting from a different place, with different business challenges and a different type of data. There is so much they must do differently.
Through extensive consultation with global businesses undergoing digital transformations, we have identified 11 steps, compiled into three broad categories, for successfully delivering value from AI projects for these digital transformers.
Build AI That Is Fit For Purpose
1. Build Trust in AI: AI must be up to the task. An AI digital marketing campaign may accommodate imprecision, but an AI built to spot when a plane engine might fail needs certainty. You cannot simply let an AI program loose on data; you need rigorous training data and training regimes. The greater the consequence of an AI error, the more rigorous the approach.
2. Don't blindly hoard data: There is a belief that more data will improve AI impact, but this is only true if it is consistent, well-tagged data. Identify the problem that needs solving, and then work smartly to identify the data best to solve it.
3. Focus on user experience: AI interaction must be intuitive or it will not be taken up. It's here that we can learn from digital-native companies: Google Photos runs neural networks, image analysis and natural language understanding, but all the user needs to master is a search bar.
4. Maintain oversight: AI is good at automating routine tasks but cannot deal with situations outside its training. To avoid AI failure, check random samples of AI outcomes against human experts and plan for expert human intervention when unexpected events occur.
Find The Right People
5. AI is about talent as much as technology: While digital transformers have different problems and need different people than digital-native companies, they should emulate the Google/Facebook approach of finding the right people for the task, not just throwing technology at the problem.
6. Mix people: AI should be designed by people who understand the problem, the underlying data and what it represents in a real-world context. The best teams include representatives from IT, operations and business teams, domain experts, AI and data analytics experts and, critically, people who can translate between these different roles.
7. Look outside your organization: Specialist AI skills rarely already exist in pre-digital organizations. Seek them out externally and embed them within business teams. Don’t limit your search to your sector. Your problem may have already been solved elsewhere.
Start Small But move Rapidly And With Purpose
8. Built momentum: Build a roadmap that identifies the business decisions that AI can inform. Focus initially on well-understood opportunities that can be executed quickly. This will build critical momentum for AI programmes.
9. Explore multiple AI projects in parallel: Accelerate this momentum by running multiple AI projects in parallel, ensuring the best ideas are progressed rapidly. This agility is how digital-native companies deliver innovation, but it's lacking in many pre-digital organizations.
10. Fail fast: Monitor your many AI projects, checking the relative performance of each, abandoning bad ideas and using successes and failures to improve training regimes.
11. Quantify value: Define measurable goals and KPIs for each new AI release (e.g., increased customer engagement, improved production line quality, reduced non-productive time). Use these to demonstrate success to financial backers and to feed back into your AI strategy.
Physical enterprises must harness the disruptive potential of AI or risk being disrupted by someone else who does it better. However, industries that are dependent on physical products or assets are starting from very different positions than from the likes of Google. Attempting to duplicate the AI strategy of a digital business is not the answer. If you follow these steps, you’ll put your organization in the best possible position to thrive in this digital world.
Written by Matt Jones, Lead Analytics Strategist at Tessella for Forbes