The Intelligence Superhighway

In the 90s everyone talked about the “information superhighway” that was created by the internet and fiber optics. AI has the same potential to become an “intelligence superhighway.” But that potential is unrealized today and instead feels like driving at rush hour.

AI is Like an Automobile

Let’s stick with transportation analogies to help explore AI’s limitations today and discover what it will take to unlock its full potential.

1. Think of AI like the automobile. 
2. Before cars, the best modes of transport were trains, horses, and feet.
3. Trains then were equivalent to today’s well-engineered systems – very expensive to build and only able to do what they were designed for. They have low flexibility and adaptability, but operate with great efficiency, speed, and scale.
4. Horses then were equivalent to today’s predictive models – more flexible in uncertain conditions than code but more likely to generate a pile of sh*t.
5. Feet then were equivalent to today’s manual processes – very slow but infinitely flexible, needed to fill in the gaps between other modes of transport.
6. Cars offered the promise of both flexibility and speed.
7. But when cars drove on roads built for horses, they didn’t achieve their full potential.
8. Today’s AI is driving on roads built for horses. 

The AI Productivity Paradox

Consider these two facts: 

1. AI can build any idea you can imagine within minutes or hours
2. Real productivity gains in the workplace remain elusive

How can both be true? See #8 above.

Case Study: AI and the SDLC

So what is the difference between a road built for a horse and a road built for a car? Let’s spell it out using the software development lifecycle (SDLC) as an example.

1. AI can do market research to develop a product strategy and roadmap
2. It can monitor user behavior and feedback to generate bug reports and feature enhancements. 
3. It can write code to solve any problem you throw at it with zero human intervention.
4. It can write tests to validate that code.
5. It can review that code for architecture, consistency, style, security, scalability, etc.
6. It can write infrastructure as code (IaC) to deploy cloud resources to host, run, and monitor that code.

Roads Built for Horses

That’s pretty much the whole SDLC, right? So why do we still have human R&D teams? 

Because even though the car is capable of driving the whole length of the road, its speed is limited by potholes and busy intersections.

1. Different people are responsible for each of the steps above today. A big group of product managers, designers, front end, backend, QA, devops, security, data engineering, data science, product marketing, client success, etc needs to have collectively hundreds of hours of meetings per week to “sync” and “plan” and “knowledge share.”
2. And they have to wait for each other to pass the work down the assembly line. Everyone has ample time to twiddle their thumbs while bottlenecks accumulate.
3. Each person is responsible for a narrow and specific aspect of the lifecycle. When your neck is on the line, it’s too risky to blindly trust AI for that aspect that you’re solely responsible for. You slow it down and do a human review while everyone downstream in the assembly line waits.
4. Meanwhile, digital systems are disconnected and walled off from each other, depriving AI of the full context and capabilities it needs to drive at full speed.
5. Permissions and authorizations for those digital systems are scoped to individuals instead of projects, so individuals using AI tools lack the authority to execute projects end to end.

People in each role are using AI and feeling more productive, but are not seeing their team overall be more efficient.  We’re still stuck in too many meetings, stuck waiting for reviews and approvals, stuck being overly cautious just to cover our asses, and stuck manually patching together disparate tools and data sources.

Building the Intelligence Superhighway

We need to build new roads and highways and learn to design new traffic patterns for AI in order to usher in the “intelligence superhighway.”

1. Accept that AI can not only do your job but can also do the job of every person on your team – yet it cannot and does not need to do all the current tasks your team does. This is subtle so I’ll reframe it for emphasis: the traditional roles in a software company may need to be redefined to unlock the full potential of AI. Don’t focus on using AI for the tasks you’ve always done – focus on using AI to orchestrate the creation of an end to end solution that may include components you are beyond the scope of your current role.
2. Give more people roles that are responsible for cross-disciplinary outputs (roles like GM and Product Manager), and hold them accountable at the system level not the discipline level. We need fewer cogs in the machine taking instructions and more drivers leading the auto intelligence to our chosen destination. Taste and judgment trump technical abilities, but attention to detail is more important than ever to keep the car in the lane.
3. We need to pave highways thru the sacred neighborhoods of privacy and security. Not every walled garden needs a bulldozer, but a company needs enough highways in the right places to achieve an optimal flow of intelligence