How to Stop People from Asking, “When Are WE Getting AI?”

What do you do if you find a mystery server under a desk that’s been running for two decades? (True story!) Or if you need guidance on how to deal with your boss telling you to use artificial intelligence (AI) instead of a human to fill your open headcount? Maybe you need to survive a blameless post-mortem when it’s actually your fault.

In the spirit of solidarity, hoping to offer a bit of humor and maybe even a few useful suggestions for IT leaders, I wrote The Engineer’s Survival Guide: Expert advice for handling workload (and work/life) disasters. The new book, deftly and hilariously illustrated by Giovanni Cruz and published by Cockroach Labs, offers top tips from experts for surviving your job, surviving the workplace and surviving whatever comes next (which, these days, could be anything).

You can get a free copy of the book on Cockroach Labs’ website. Meanwhile, enjoy this bonus survival tip (director’s cut version!) we resuscitated from the cutting room floor.

How to Answer, “When Are We Getting AI?” without Screaming Inside

Ah, AI — the trend you, as an IT leader, must embrace, lest ye be rendered desperately obsolete. This is one rabbit hole from which there is no escape.

For a long time, generative AI was “coming soon.” AI was out there, hovering vaguely on the horizon, even if there was no way to know when it would get here — or what it would look like when it did. Early AI felt more like a party trick than a useful tool.

Then suddenly AI got here all at once, bringing with it fast-follow, sky-high pressure to incorporate AI into your product, service or software ASAP. FOMO or no, when you’re tasked with bringing new tools and tech into your organization, these days people have just one question: “When are we getting AI?”

It comes at you nonstop and from every angle: at the coffee machine, in the elevator, in sessions with your remote team, in the restroom. Even the lunchroom is not safe: You want fries with that? Hey, by the way, when are we getting AI?

It’s easy for them to ask; they’re not the ones figuring out a safe course between the twin perils of wasting resources by bringing in AI too soon versus waiting too long and lagging behind competitors. This is achingly new tech, and AI’s risks are real, including discrimination, violating consumer rights or (mis)leading organizations into making costly bad decisions.

Looking for too long before you leap can quickly turn dangerous when enterprises freeze in an infinite loop over everything they don’t yet know about AI. At the other end of the spectrum, though, lies too much action — too fast and, quite often, in the wrong direction. (For example, does the world really need AI-powered toothbrushes?)

Your job is to safely pilot the SS Enterprise between falling into the black hole of analysis paralysis while not flying too far and too fast without enough data, straight into a supernova. The trick is to focus on the things that won’t change — yes, there are some — even as the generative AI space morphs around us faster than tribbles reproduce.

Via GIPHY

Doubling down first on the non-fungible fundamentals leaves you ready to launch when the stars do align for bringing the right AI tools (or cyborgs) into your organization. Even better, it gives you a plausible yet positive way to respond when asked, for the eleventy-billionth time, about the plan for AI adoption in your organization.

Follow these survival tips and eventually you, too, will be able to say a confident “Good morning!” to your colleagues without fear.

New Tech, Who Dis?

Architects, senior engineers and others in positions of IT influence must resist the black-hole gravity field created by the pressure to chase shiny new tech (AI, AI, AI! Did we mention AI?) that comes from every angle.

No matter how many paradigms AI is pushing, it still requires the same evaluation process as any other tech your organization adopts. This immutable fact is your ace in the AI hole. Everybody wants the shiny thing, but few want to do the serious upfront analysis to make sure it’s the right thing: Adoption cost. Capability. Usability. Interoperability and ease of integration. Potential threats and risks. And let’s not forget everybody’s favorite, legal compliance.

So, when people are pounding at your door (proverbially speaking — who has an actual office anymore?) and demanding AI answers, you can drop this line on ’em: “We are currently in the process of aggregating proactive evaluation infrastructures to disintermediate the known unknowns before integrating AI into our mission-critical operations.”

This is just a fancy way of saying, “We are evaluating AI technology for adoption.” Then, you tactfully slip out the aforementioned door while they are working out what you just said.

Off to Work, You Go

Cool. Now that you’ve bought yourself some breathing room, here is where the real work begins to answer their demands and bring AI into your organization. You will need quality data, APIs that enable interaction with your business’ capabilities and some hands-on experience with the types of APIs in the generative AI ecosystem.

1. No Data, No AI

The first thing to consider is your data foundation since any AI you bring in will need to train on that data. GIGO still applies. Are your data management practices sound and thorough? Is your data well-cataloged and good quality? Do you have metrics that track data freshness and completeness?

In other words, don’t feed your AI dirty data. If you don’t have a good answer for each of those questions (don’t feel bad, just about every organization falls short in at least one of these areas), you’ll need to do a proper data cleansing — preparing data for analysis by fixing or removing anything incorrect, incomplete, irrelevant, duplicated or just plain old improperly formatted.

2. Practice Makes Perfect

More knowledge about and experience with AI leads to making better and more informed decisions. (ChatGPT made me say that.) So, learn by doing. Pick an AI platform that intrigues you and play with the tech. Pick a useful use case. Pick a large language model (LLM). Pick a database applicable to your use case. Then do some stuff.

AI technologies will inevitably change — and quickly — but the patterns for interacting with them will not. (At least until we reach artificial general intelligence when the machines can reason and make decisions just like a human. Then all bets are off.) Most of the time it’s just interacting with more and different APIs.

3. Speaking of APIs…

An API defines the interactions possible between separate and distinct pieces of software, so they can request and return or present services — all without one needing to know anything about the other. This layer of abstraction enables simple execution of complex interactions, which to be honest, might be the only simple thing about AI.

APIs allow an enterprise to better deliver diverse data and services to customers, both internal and external. Whatever AI tools you adopt are almost certain to be your biggest internal “customer.”

This is why the time to assess your API foundation is before bringing in AI. How well are your organization’s business capabilities served using functional, reusable enterprise APIs? To harness AI — particularly for business goals like automation and capturing efficiencies — you are going to need a way to interact with it, both digitally and physically. APIs are how to open the pod bay doors.

https://getyarn.io/yarn-clip/f24a6ca9-5c66-4782-a672-ea06a56464e8

Via Yarn

4. Robots Need People, Too

Great! Now that you’ve established a solid foundation of AI-friendly data quality and management, created APIs for any necessary AI interactions with your existing system’s capabilities and services, and gotten yourself some hands-on experience with the generative AI ecosystem, you are ready for the final step in onboarding AI: recruiting the human support crew.

In fact, even while working through the first three phases of AI testing and exploration, you are probably already building relationships and collaborating with other teams as part of the AI adoption initiative. The data and analytics folks will help with the data cleaning. Engineering, platform and developer teams will help wrangle the AI APIs and try them out internally. If you haven’t already, now’s the time to include your security, compliance and (of course) legal teams.

Choose Your Wormhole Carefully

Yes, there is a lot to learn about harnessing AI to provide real business value without, you know, bringing about utter destruction. And, yes, the AI ecosystem (not to mention the entities that bring it to us) is constantly evolving. The key is to set aside everything that is currently unknowable — nobody knows this stuff; we are in the process of inventing it! At the same time, don’t let the lure of tantalizingly shiny, truly paradigm-shifting technology (not to mention crushing peer pressure) cause you to dive head-first into what could very likely be the wrong AI wormhole.

Instead, focus on the things that won’t change and that are within your control. This leaves you well-positioned to take advantage of the right opportunities for your organization when time and conditions are right.

Congratulations! You have aggregated the proactive evaluation infrastructures and synergistically orchestrated error-free human catalysts. Your executive leadership will thank you, and you can finally walk to the bathroom in peace.

If you’ve read this far and found it to be entertaining, possibly even useful, go read the actual book (for free)!