Sten Vesterli's Blog

Making IT Live up to its Promise

Tag: AI

Another Place Not to Use AI Chatbots

IT Leadership, , ,

Alaska wanted an AI chatbot to give legal advice. Anybody care to guess how that went?

Yes, not well. They are now 15 months into a 3-month project, but expect it to go live this month.

I’ll make a prediction: This project will end up in the bucket of IT projects that sink without a trace, leaving only cost and no business benefit whatsoever. Sadly, one in four IT projects still end in that bucket. My hunch is that this number is increasing, as AI is incorrectly applied to more and more use cases.

If you are going to implement AI with an LLM, don’t do it in a critical application like giving legal advice to citizens. There is no way to stop an LLM from hallucinating. The only way to automate advice and be sure it is correct is with old-school expert systems.

Good AI Advice

IT Leadership, , ,

Who advises you on AI? Don’t say ChatGPT. Also, don’t take advice from random blowhards on LinkedIn. You need advice from someone who has a realistic view of your situation and your business.

As a consultant, I’m all for having external advisers to bring you an outside perspective. But it is equally important to have a well-founded inside perspective.

I recommend establishing an internal AI advisory board within IT. Appoint some people interested in AI and give them a modest time budget to keep up to date with what is happening in the field. Make it as diverse as possible – juniors, seniors, developers, sysadmins. If you are fortunate enough to have diversity in gender and ethnicity in your IT organization, also utilize that. Have your AI board meet with the CIO/IT leader regularly, and also have them present at department meetings.

Inside people are much more invested in finding AI tools that can truly help in your specific situation. They are also the ones who will suffer if you implement bad AI. That gives them a very good ability to see through exaggerated vendor claims.

The Right and the Wrong Way to Use LLMs

IT Leadership

There are two ways to use Large Language Models. One works well, the other much less well. Over the holidays, I’ve been talking to a lot of family and friends about AI, and it turns out that many people conflate these two approaches.

The way that works well is to use LLMs deductively. That means starting with a lot of text and distilling some essence or knowledge from it. Because you are only asking the AI to create something from a text you have given it, it has much less room to run off on a tangent, making stuff up. At the same time, it can show off its superhuman ability to process large amounts of text. In an IT context, this is when you give it dozens of interlinked files and ask it to explain or find inconsistencies or bugs.

The way that doesn’t work well is using LLMs inductively. That is when you ask it to produce text based on a short prompt and its large statistical model. This allows it to confabulate freely, and the results are hit-or-miss. In an IT context, this is where you ask it to write code. Sometimes it works, often it doesn’t.

Whenever you discuss LLMs with someone, set the stage by defining the inductive/deductive difference. If people already know, no harm done. If they don’t have this frame of reference, establishing it makes for much better conversations.

Learning From People, Not From Documents

IT Leadership, , ,

Implementing AI has a critical and often-overlooked problem that Raman Shah just reminded me of in another discussion: It can only learn what is documented.

When you teach an AI to perform a process in your organization, you train it on all the internal documents you have. But these describe the RAW – Rules As Written. They do not describe the RAU – Rules As Used.

It takes a lot of care to do process discovery properly. You need to have a human being interview the people doing the work to find out how business is actually done.

Work-to-rule is a classic form of industrial action where workers do exactly what they’re told in order to stop production without going on strike. If you train an AI on just your documents, you are asking it for a work-to-rule stoppage.

Break the Law More?

IT Leadership, , ,

Do you want your AI to follow the rules? That’s not as clear-cut as it seems.

You do want your AI to follow your instructions to not delete files. The Google Antigravity “vibe coding” tool achieved notoriety this week after wiping a developer’s entire D: drive instead of just clearing the cache. It said, “I am deeply, deeply sorry.” Google has promised to look into it.

On the other hand, Waymo self-driving cars in San Francisco have been notorious for meticulously following traffic rules and blocking traffic. Waymo just updated its software, giving its cars more of a New York taxi driver attitude. They no longer block traffic, but on the other hand, they now do illegal U-turns and illegal rolling “California stops” at stop signs just like humans.

Before you start getting a computer – whether deterministic or AI – to do a process, make sure you understand both how the process is documented and how it is actually done today.

AI Agents are a Stupid Idea

IT Leadership, ,

AI agents are a stupid idea. Consider that we’ve had the option to program deterministic agents to handle most of the suggested AI workflows for a long time. Yet we didn’t do it? Why? Because it turned out to be hard. We did not have good data or good APIs for the actions we wanted to take, and we always encountered lots of difficult edge cases.

Yet the AI vendors are proposing that we now try again. This time with stochastic algorithms that we are never quite sure what will do.

Agentic AI means that we take a problem that we could not solve with deterministic programming, and add another problem on top. And that is supposed to be the future? I don’t think so.

Are your AI Systems Insurable?

Uncategorized,

It will not be technology or regulations that limit AI; it will be insurance. Several major U.S. insurance companies are petitioning lawmakers to allow them to completely exclude AI risks from their coverage, arguing that they pose an unmanageable risk.

Insurance companies are already fighting with Air Canada over who will pay for the fictitious discounts that their chatbot invented and they had to honor. Next time something like that happens, AI will allow thousands of customers to easily create plausible claims, but the insurance company cannot use AI to handle them. This asymmetry has them scared, for good reason.

If you are running AI systems, take a good look at your Tech E&O insurance. You are likely to find that it already limits coverage from some types of AI incidents. When it comes up for renewal, you will find more AI excluded. Insurance is just as real a limitation as regulation and technology.

Intentional Analysis

IT Leadership, ,

Isn’t it funny that the only people saying AI will take over the world are those selling the stuff? Of course, supported by the usual coterie of consultants looking for a gig, academics looking for attention, and clueless journalists looking for a sensationalist headline.

When I was in high school, we learned to do intentional analysis – considering what motivations the author of a text might have. That skill seems to be widely forgotten when discussing AI.

It also applies to programmers dissing AI as glorified autocomplete – they have an interest in telling everyone that they are still indispensable.

Elle King sang that there are “always two sides and the truth.” It is your job as an IT leader to look at the messengers on both sides, evaluate their claims and credibility, and figure out approximately where the truth lies.

Blocking AI is an Unwinnable Battle

IT Leadership, ,

Using AI is not cheating. It is a way to become more productive. You pay your employees because they perform tasks that create value for the organization. So it makes sense to let them use the best tools available to do their jobs.

Just like some schools are trying to prevent students from using AI, some companies are trying to outlaw AI. It won’t work. Research shows that 47% of people who used AI tools experienced increased job satisfaction, and 78% were more productive. You can’t fight such dramatic numbers with a blanket prohibition. If you try, your employees will use AI on their phones or in an incognito browser session while working from home.

By all means create rules about how and where employees can use AI, and explain them thoroughly. But trying to ban AI is futile.

Business Knowledge Beats Technical Skill

IT Leadership, , ,

Most of the value of an IT developer comes from his knowledge of the business. His knowledge of specific programming languages or tools comes a distant second. With AI-supported development tools like Copilot, this value balance becomes even more skewed towards business skills.

That’s why I’m appalled every time I see yet another company replacing hundreds of skilled IT professionals. I’ll grant you that some organizations have too many people and might need to trim their headcount. But often, organizations are trying to kickstart a digital transformation by replacing old hands with a crowd of bright-eyed young things with the latest buzzwords on their CV.

Getting a new crew with better tools and techniques means you can build software faster. But by getting rid of the experienced people, you lose your ability to build the right software. Moving slowly in the right direction beats running fast in the wrong direction.