(Pixabay)

I haven’t been bowled over with the AI initiatives from software vendors and customers. The lack of comprehensive planning by users is troubling and wasteful. The vendors seem focused on creating AI platforms and creating incremental improvements to their existing software modules. These kinds of actions suggest that AI strategies may either be missing or misaligned.

There’s a technology maxim, Fubini’s Law, that I first saw in the 1990s. It succinctly describes how people and businesses see and utilize new technologies. What it described were four or five steps businesses would follow as they came to understand and deploy new technologies and then use this knowledge to re-think/re-imagine their business, its processes, etc.

This Law remains spot-on even to this day.  Unfortunately, companies, in their zeal to slap AI on everything, don’t remember the past and are doomed to repeating past mistakes.

Let’s examine what firms could’ve/should’ve done differently as they got their feet wet with AI. What we’ll see are wasteful strategy mis-steps, missed opportunities, and more.

Fubini’s Law #1 – people initially use technology to do what they do now, but faster .

We’ve seen this play out across many successive waves of new tech. The initial computer applications of the 1950s-1980s helped firms automate all kinds of computationally and manually intensive tasks. These technologies calculated payrolls, printed invoices, etc. They absolutely allowed firms to replace a number of clerks with systems. Those early automation efforts may have relied on really old-school tech exclusively with mainframe computers (e.g., punch cards, computer service bureaus, magnetic tape, etc.) but the new speed, accuracy and productivity they triggered started to positively impact nations’ productivity statistics. And yet this initial wave only really helped the very largest organizations: entities that could afford custom code development and mainframe computing.

Subsequent automation waves included:

The rise of mid-range/mini-computers and early application software packages. This wave brought new tech to a much larger population of businesses, the mid-market, while triggering productivity and efficiency boosts to a huge number of businesses globally.
The mass uptake of personal computers. The first personal computers opened up new capabilities to businesses and business people everywhere. These tools and the new application software (e.g., WordPerfect, Lotus 1-2-3, PageMaker, etc.) made planning, budgeting, presentation development, reporting, etc. highly democratized and inexpensive.
The availability of web browsers, networked computing and the World Wide Web. This wave showed how different firms could better communicate, coordinate and interact with each other. Communication costs plummeted, reaction times shrank dramatically, visibility in supply chains was materially enhanced, etc.
Several others waves, like the ecommerce wave, the cloud era, and, today’s AI age.

It’s absolutely essential to note that the biggest productivity gains that companies have reaped via automation likely occurred in the past – the distant past to be sure.  In fact, holding all other factors constant, it’s the early waves of applying technology to a manual process that triggered massive productivity gains. 

Conclusion – your first strategic question to ponder is, how confident are you that your firm will find huge, net-new, productivity and efficiency savings via AI that weren’t already realized in part or in whole from previous automation efforts?

Fubini’s Law #2 –  then they gradually begin to use technology to do new things.

Remember when cell phones got a digital screen? That screen meant that users could also see and send text messages. They could also take and view photos. They could even play video games on their phones. The cell phone wasn’t just a telephonic device anymore. That new generation of cellular technology allowed users ‘to do new things’ and did they ever do it.

In today’s AI Age, this second stage of Fubini’s Law has taken an interesting turn. It’s not application software vendors that are leading here. No, this honor, if you will, is reserved for the miscreants who use citizen-AI tools to:

Overwhelm HR recruiting systems and processes with record volumes of resumes and applications, many of them with dubious value or legitimacy
Make any resume or application exactly match the keywords and experience level that employers desire
Create highly believable but fake T&E expense receipts
Create near perfect fake invoices that companies incorrectly pay
Create fake voice mails, video calls, etc. that mimic a company’s top executives

Ironically, what is happening here is actually subverting the first step in Fubini’s Law. These AI creations are actually making fraud, criminality, etc. occur at a blindingly fast pace and at extremely low cost but the results are not improving productivity or efficiency. They are doing the opposite.

Application software vendors can’t point to a lot of net-new capabilities that AI is helping deliver. They can point to some small successes here including AI tools that help generate succession plans, drafts of performance reviews, financial briefing packets, journal entries and other back-office deliverables. Many of these ‘products’ are simply an AI enabled capability of a pre-existing process or application. These aren’t really net-new ideas with most being an AI-assist applied to an existing application.

Another example of low/no value-added AI work occurs when businesses use AI to build an entire application (e.g., payroll or accounts payable) that replaces a solution the firm had licensed or subscribed to from a software vendor. Yes, the new AI application might be developed quickly but it simply does (only some of) what the purchased app did. It won’t necessarily operate all that differently or create significant amounts of new value. It might (emphasis on the word ‘might’) be a less expensive solution to acquire and possess in the near-term. However, take it from someone who’s built and reworked numerous Payroll, Financial and other applications, creating and supporting application software (even if it’s AI generated) comes with an expensive, long-lived tail. And, the more regulated a process is, the bigger that tail gets.

Conclusion – for now, the second part of Fubini’s Law seems mostly AWOL.

Fubini’s Law #3 – the new things change life-styles and work-styles

A number of pundits have spent the last year or two pontificating about how AI will dramatically re-shape the nature of work, jobs, management, etc. Some pundits seem to have highly pessimistic views while others speak of AI’s impacts on the workforce in almost giddy language. So far, no single perspective can be viewed as correct or accurate. It’s just too early to tell how AI will change work, let alone how related skill, job and work needs will impact disposable income, wealth creation and other factors that impact a person’s life-style.

Pundits have some vision, though. They already foresee a future where management ranks will be thinned out (a la delayering). Low-skill clerks will no longer be needed but highly skilled ‘person in the middle’ subject matter experts will be. Some industries may be highly impacted with many speculating that consulting and IT services work (e.g., managed services) will be adversely impacted.  AI (via AI chatbot functionality) is already impacting call centers.  These changes are, for now, still nascent and not well understood or documented yet.

What can be said with some certainly is that AI will impact work, people, businesses and society (to pick just four items). How it impacts everything is simply a big unknown for now.

This ambiguity is due in part to a significant lack of effort and imagination on the part of many AI users, developers, etc. Significant changes to people’s life-styles and work-styles won’t happen when people are content to create small incremental AI utilities, tweaks to existing applications, or, look-alike/work-alike AI processes that mimic existing systems.

The third part of Fubini’s Law seems pretty elusive right now. For this part to manifest, we need to see development teams ideate all-new concepts, work designs, etc. in bold, daring, novel ways. Work must be radically reimagined and reinvented.

AI developers should have tools to help them document as-is processes and ideate new to-be workflows and process designs. AI-relevant ROI calculators should also be part of the build out toolkit, too.

Conclusion – a lot of work is needed before the third part of Fubini’s Law becomes a reality.

Fubini’s Law #4 – the new life-styles and work-styles change society

We are in the very early days of these changes.  Some clues are already appearing though.

Stock valuations may be a great early/leading indicator. Stock prices of traditional application software vendors, industry analyst businesses and professional services firms are way down these days as investors try to figure out how many people these firms will actually need in the AI age and whether customers will want to use these service firms at all (i.e., use AI instead for much of their future needs). Those are some big question marks. They suggest that investors believe a material change in these firms, the nature of the work they perform, etc. is afoot. They may be right but by how much is unknown.

Some pundits are expecting changes appearing all over the employment/work landscape with some predicting relatively dire employment prospects for large swaths of the workforce. In the bleaker examples, there are discussions that governments may need to exercise more input into AI-led job losses and/or the need for a minimum mandatory universal wage for people.

At least people are talking about these potential situations and the issues we’ll likely see. The solutions seem quite thin for now.

From a technology perspective, how people interact with software applications will be changing dramatically. We’ve had several generations of apps UX. I’ve personally worked with a punch card interface, simple menus, PF keys, spreadsheet input, 24X80 green screens, menus, and cell phones. Now, for the first time in many, many decades, the user experience (UX) will likely be dramatically replaced. This time it will be by an AI prompt line. AI agents will not only handle a number of predictable, repeatable tasks without the need of a screen or menu, they will also alert subject matter experts to anomalies, new business (and external) events, etc. We will transition from a world where work involved the processing of transactions to one where AI predicts problems, proactively suggests solutions to all manner of events not just transactions that directly impact the general ledger.    

At the same time the audit trail will soar in importance and become a key part of the new UX. Why? Users and subject matter experts will need to know how an LLM or agent came up with its findings and decide if this probabilistic response is actually viable and valuable. 

Conclusion – while workers rarely gave audit trails much thought with traditional systems, they will in the AI age.

Fubini’s Law #5  – and eventually change technology

This is the really interesting part of Fubini’s Law. When a new tech (and all of the enabling tech that comes with it) finally catches on, then all kinds of new ideas, new usages, etc. will blossom and these will trigger innovators, entrepreneurs, etc. to explore a new wave of technologies. For example, the introduction of smart phones triggered all kinds of employee self-service applications, mapping software, online banking, social media applications, and thousands of other solutions. What started out as a phone morphed into something entirely different. Today, few would want a basic device that just makes phone calls. The same thing happens with business technology, too. It’s not just that the original technology breakthrough (e.g., telephone) was deficient, it’s just no longer market relevant given newer options.

Right now, AI innovation is limited. It’s limited by a lack of complementary technologies (e.g., the presence of hundreds of thousands of AI Agents would likely spur on all kinds of process innovations) and ideas/inspiration. Libraries of thousands of pre-developed algorithms (along with key reference data) could launch all kinds of amazing forecasting tools that would make supply chains, budget plans, inventory management, manpower planning, transportation network planning, production planning, etc. so much more efficient.

In thinking about the fifth part of Fubini’s Law, one should look at the other tech, data, resources, etc. that could dramatically supercharge the impact that the current innovations could produce. Those capabilities might not be widely available or economical now but that could change quickly. And when it does, it creates all-new kinds of solutions that may completely change the economics of today’s AI age. Since we’re in the very early innings of the AI age now, it’s hard to see examples of what’s to come. But, at the same time, we should be hearing of the far forward-looking concepts leaders have as they look to make all-new solutions with AI being part of the mix. 

Conclusion – you won’t get these breakthrough ideas if your head is all wrapped up in incremental improvements.

Your strategy to-dos

The Financial Times had an excellent piece on AI adoption and, when reading it, one can’t help but realize how many missteps (e.g., tokenmaxxing) companies/executives have made to encourage AI usage in their firms. When I read this, I was struck with how little thought was going into creating a great AI strategy and how many firms were just hoping that poorly thought-out directives would somehow lead firms to AI greatness.

A great AI strategy doesn’t happen accidentally. It doesn’t happen by one executive’s whim/command either.

Given the above, it’s time to have a thoughtful, premeditated AI strategy. Here are just some of the strategic concerns re: AI and application software that companies should consider:

AI innovation is still occurring. How long will the AI interest in ML (algorithmic), Agentic and Generative AI last? When will AI possess great context capabilities? Ditto for General Intelligence and World View solutions? Should we place more attention on the AI capabilities to come versus those already present?
How do we roll out AI solutions when the costs are highly volatile and range from free to exceedingly expensive? What’s our process for identifying the ROI of individual AI initiatives? How will we know we are working on the AI initiatives with the greatest return (and not just spraying AI all over the enterprise with little rhyme or reason)?
Do we know how LLMs (Large Language Models) will evolve? Should we bet on narrower, smaller and cheaper LLMs vs. expensive large, public, general purpose LLMs?
Some AI developments will require net-new training on tools, reskilling people involved in affected processes, new metrics, new analytics, etc. Does your strategy/plan identify all of these and the other change management needs (e.g., stay pay, reduction in force costs, etc.)?
Has your firm researched where its AI services will be hosted and what kinds of environmental impacts (e.g., water consumption, electricity usage, noise, etc.) these specific centers will have on local communities? Will these generate adverse publicity for your firm? Where is this assessment?
Workers impacted by AI include more than those who process transactions. Have all new ‘person in the middle’ roles been identified and the needed skills located? How will your firm permanently and accurately capture the tacit knowledge that experienced (but soon to be laid off) workers possess?
What pre-requisite work should we focus on now? Do we have accurate, complete, well-defined workflows/processes? Is the data we need clean, available, accurate, normalized and at a workable level of detail?  Have we trained everyone as to the new risks AI tools can present? How reasonable are your estimates of the lifetime support and usage costs of all proposed AI initiatives so far? Whose budget does this hit?
A good strategy has a well thought out exit in mind. When will today’s new AI powered apps/agents in your firm likely hit EOL (end of life)? Who will determine this? Is this EOL part of the ROI assessment? What are the planned support and upgrade costs for all of your AI developed products? What budget are these in?
Does your strategy assess the need for newly developed countermeasures of citizen-AI based threats? How else will your firm address T&E fraud, jobseeker fraud, executive impersonations, synthetic persons, invoice fraud, etc.?
Do we have a timely, complete and effective process to assess newly ideated AI initiatives? Are the suggested initiatives big and bold enough to develop outsized ROI? How will you decide which initiatives do/don’t make the cut? What are the prioritization criteria?

My take 

There are several pundits and software executives who are guilty of excessively fanning the hype as well as the doom/gloom predictions of AI. And, there are others who are piling on with cheery uplifting visions of an idyllic post-AI life. What all of these groups share is a material lack of critical thinking and an absence of great data to back up their opinions. Both groups also fail to look at the events of prior technology innovation cycles. I actually have some choice names for these folks but then, this publication is a ‘G-rated’ news source.

That said, it’s up to software buyers and users to do a better/great job of assessing the AI world and its potential to economically impact their firm. Vendors and pundits won’t as they are too invested in the past and/or busy trying to keep their old code market relevant.

Finally, have some fun folks. If an integrator or vendor comes pitching your firm some half-baked AI solutions, call out their incremental, non-strategic thinking and dare them to show you anything visionary beyond agents, algorithms and generative AI reporting packets. They will probably fail to do so and in that moment expose themselves for being the poor thinkers and market watchers that they are. Your firm deserves great insights and experts not hacks.

For me, I don’t suffer fools, gladly or otherwise, and neither should you….