The change to AI is going to be quite hard

Is this a mailing list?

Sort of. Let me say up front that you can reply “go away” or click the unsubscribe button and banish this email forever with my apologies.

But let me explain what “sort of” means.

You are a leader who I have discussed the topic of AI with in the recent past and I’ve enjoyed our interaction and your interest. I have the privilege of talking to dozens of communicators and marketeers about AI, as well as some of the best technical minds in the industry. Plus my job means that I get to think about AI 24/7*.

This has given me some unique perspectives, insights and thought provoking questions about the way in which these two worlds (AI, and communications / marketing) are heading. I would like to share these thoughts** as I hope they will be useful and informative as we grapple with the fundamental change that AI represents. I’ll also apologize up front for sharing in a long-form format - I’d like to move above the noise of constant AI commentary and take the time to get more deeply into some subjects and give them the proper amount of thought. Serious topics cannot be dealt with quickly. Unfortunately that means a lot of words.

Why not just share all this on LinkedIn? Because, basically, social media is blah. I can understand why newsletters have become so popular - they allow for a focused and personal discussion away from the noise and grandstanding of social media. So this is my intention, to somewhat irregularly share with you - as leaders in the communications & marketing field - some of the insights I have the fortune to observe as we head into this new world, and to hopefully create helpful dialogue in the process, and perhaps leverage our collective minds along the way.

_{*But I won’t be mentioning the thing we build as part of my job here - this is just me and my thoughts.}

_{**Obviously absolutely nothing confidential is ever shared}

This week…

The change to AI is going to be quite hard

“AI at work is here. Now comes the hard part” reads the introduction to Microsoft’s recent Work Trend Index report. It paints a challenging yet familiar picture summarized with the following words on the front page:

There is pent up demand and need for AI solutions in organizations, but at the same time there is a reluctance or a hesitance to commit to implementation. Again quoting from the report:

This is put into more concrete terms with some statistics: 79% of leaders agree their company needs to adopt AI to stay competitive while 60% of leaders worry their organization’s leadership lacks a plan and vision to implement AI.

The discrepancy between these two statistics is alarming.

Yes, change is hard

Change is also not new, people in any organization deal with change all the time. However, there are some fundamental factors which make this change fundamentally different and more challenging.

It’s worth having a look at these factors in some detail:

In 2022 ChatGPT changed everything overnight, being the fastest technology in history to reach 100 million users.  It achieved this because firstly it is incredibly simple to use being based on the most natural user interface of all - human language. And secondly because it can be used to do pretty much any knowledge task you ask of it (actual mileage may vary). “For the first time computers understand us instead of us having to understand computers” says Satya Nadella.

This makes AI (or specifically large language models) the first technology to be truly universal in the business sense, and so suddenly everybody from the top to the bottom of the organization is asking about the AI plan, albeit sometimes from a standpoint of opportunity and sometimes from a standpoint of fear or concern.

So whereas historically technology has initially arrived in the form of very specific applications for very specific problems and thus been easy to delegate, AI is perhaps the first technology change that many senior leaders are directly confronted with as having a very direct impact on their organization.

IT and AI

In many (but certainly not all) organizations IT was caught off guard by AI. Suddenly here was an application that everybody was using via their web browser, representing perhaps the biggest shift in technology since electricity, and IT were completely uninvolved.

The first reaction was to play the governance card and assume a gatekeeping role, in many cases blocking the pesky ChatGPT website. Then, fortunately, while IT was contemplating their next move Microsoft (who deserve massive credit for their early and very large bet on AI) came to the rescue with a host of AI solutions, most notably CoPilot which IT are now able to offer as their solution and answer to AI across the organization.

The result was that six months ago many firms (again, certainly not all) claimed their answer to AI was to delegate the problem to IT and that they would experiment with CoPilot or another Microsoft solution. But today, an increasing number are finding this solution falls short and I’m hearing people express disappointment or disillusionment as this approach does not look like it will usher in a complete new world of AI driven benefits.

This might be for the very same reason that lead to ChatGPT’s success: Chat based large language models are generic productivity tools that are immediately impressive, broadly very helpful but often specifically quite useless. AI is a data intensive technology and people are learning that getting value from AI is dependent on it being integrated into the problem space and its relevant data.

The analogy I often use here is that whereas Excel allows us to do pretty much anything (with a bit of work), we still rely on very specific applications like SalesForce that are designed to solve specific problems. ChatGPT is the Excel of AI.

The apparent ease of use of a chat based technology might also have deceived us. A strikingly large number of human beings are very bad at asking questions and even worse at giving precise instructions. This possibly makes chat / natural language the worst possible interface to AI.

The lesson here is that AI is not just ChatGPT, or CoPilot, or any generic solution. And the not so subtle question is: is the IT team the group that can figure this out and deliver the very specific solutions that you are going to need as communications and marketing? And when we get into problems of how organizations will inevitably need to adapt hand in hand with AI (41% of leaders expect to redesign business processes from the ground up with AI) we are suddenly far away from the competencies of most IT teams.

I don’t want to dunk too much on IT here, however. Because what I’m guilty of doing in the preceding paragraphs is probably the first root cause of why this change appears so hard: making AI seem harder than it actually is.

AI mystification

In many conversations there is a level of mystification or sorcery or voodoo magic associated with AI. What crazy science goes on behind it? “Let’s send everybody on a prompt engineering course so they can spend hours tinkering with text output.“

Confounding this situation is the extreme noise level, the proliferation of solutions and a constant stream of not very well informed media coverage. (one of the reasons I wanted to start a regular email update, but please tell me if I’m only making things worse).

And there is little outside help available. Aside from some notable (and excellent) exceptions many agencies and consultancies are still busy figuring out AI, what it means for them, and how it will change their own business models. There seems to be more internal discussion than customer solution at this point.

However there is some hope on the horizon. Solutions are evolving and becoming more problem focused. Product managers are slowly moving from “let’s put AI everywhere and call everything AI” towards asking the fundamental question “what are the problems” and solving these problems with straight forward solutions.

A matter of time

The second and most significant root cause is that while AI might not be so complicated, change does require time investment. The above statistic confirms what I hear regularly - teams are 100% engaged in keeping up with daily business without the time to step back and rethink. Communications never stops. Ever. Which is an unfortunate dilemma because AI is perhaps the ultimate solution to team overload.

I don’t think that there are many helpful answers to that problem. Other than my penultimate  quote from the report:

This suggests that we need to frame the problem differently. The problem is not implementing AI, but safeguarding the human intelligence in which we have so much invested. And AI is a key part of the solution - 90% of users say that AI helps save them time, focus on their more important work (85%) and ultimately enjoy their work more (83%).

Rather than looking at AI as an end in itself, or a solution looking for a hard problem this could be the real solution-problem pair that we’ve been missing and which helps AI get a proper head start in organizations - or at least the thing that makes the hard part worthwhile.

Whenever it comes to such analysis I believe the aim is not always to be right, but to be less wrong. I’d really like any feedback or alternative viewpoints you may have. My goal for this newsletter is to create dialogue and collective problem solving. Most certainly some things here will be wrong so please do challenge me so that we can evolve our ideas and understanding through future editions.

Lastly, please forward or share with anybody else you think may be interested. If you were sent this email by a friend, please consider subscribing for more.

Thanks,

Mark.