Minor updates & clarifications added 9/13-17/23
You need to read this. I mean it.
Yes, it’s long (9 minute read, ~30 with linked docs), but so are your days of work now totally impacted by the emerging superintelligence penetrating all facets of business and life. Summaries often fail to convey actionable understanding, but time is precious, so here’s the gist:
“With broad exposure to many topics, AI/LLMs [Large Language Models like ChatGPT] have some potential for interdisciplinary synthesis if explicitly prompted or guided to draw connections between disparate information. However, without external direction, LLMs still lack the innate ability to autonomously integrate insights across datasets in creative and synergistic ways. ( . . . ) Even as LLMs grow more advanced, harnessing their capabilities will require human-AI collaboration” (Claud 2). “Human leaders will retain an advantage for at least 3 years and probably between 5 to max 10 years” (paraphrased from three LLMs’ answers recorded here).
“Once AI reaches the level of AGI (Artificial General Intelligence), it will, by definition, be able to perform any intellectual task that a human being can” (ChatGPT).
“This essay outlines strategies that consultancies must start now. All other managers need to start planning now” (a human, me, who stubbornly generated 100% of this post’s non-quoted text!).
Managers are scrambling to understand LLMs and their impact on business life and bottom lines. They should. Using search marketing as a lens to focus on any business, this post provides a conceptual understanding of and strategies for AI/LLM’s strengths and managerial limitations. Specific tools and procedures will be spelled out in separate documents for DISC’s clients.
DISC had our version of Google’s “Code Red” when ChatGPT rocked the world several months ago. Each DISC person produced a report leading to my master report that feeds this post. DISC’s model of the 3-years Entity/MUM revolution in search marketing remains intact, but some procedures for harnessing that model have been transformed by AI. I challenged the top 3 LLMs to undercut the thesis of this post, and all three supported it. I strongly encourage reading the bolded parts of that conversation referenced below, wherein the LLMs deliver enduring comprehension of their nature–with appalling intelligence.
What qualifies me to write this? I address this more at the end. The human brain processes about 1016 synaptic transmissions per second (1,000,000,000,000,000), and I’m not in charge of any one of them. Thus I claim no bragging rights for my gift, paid by deficits in other areas, for seeing the evolution of complex systems. My late 1990s writing about the direction of search engines proved true over two decades, and I launched DISC into search marketing before Google existed. For about three years now clients have heard me propounding the importance of integrations among web marketing services: Here, now, you’ll grok why.
A 2-Dimensional Model for Understanding the Integration Imperative
Think of an image of a brain’s 3-D neural network, where an input like an idea spurs an electrical web of connections throughout the brain’s parts over a brief time that adds the 4rth dimension. Let’s reduce that to 2 dimensions. The LLM takes the request or “prompt” as a single point, and outputs a line extending left and right by stitching information from its database’s content related to the prompt topic (hence Google BERT’s “Bidirectional Encoder Representations from Transformers”). LLMs do this stunningly well. Now add parallel lines representing related topics or fields of work, for example a line each for SEO, PPC, CRO, ROI Reporting, and client sales data concerning relative profitability of kinds of leads. Superior results come from connecting those horizontal lines with vertical lines at key points. Those vertical connections flow strategies and tactics from one horizontal line to the other, enriching each with one another’s AI-assisted human labor, resulting in a synergized mind-map of strategy. LLMs are not good at knowing where (and when, over time) to draw those vertical lines, and this is why human professionals’ integrations remain essential. I envisioned this dynamic years ago when I produced DISC’s Brand Values statement: “Innovation Efficiently Integrated Is Integrity Proven by Results.”
In short, LLMs draw great horizontal lines; humans bend and connect those lines within mind-maps encompassing how your firm adds value.
Strategies for Succeeding in an AI/LLM World
Prospering will require two main processes:
- Reduce operations costs by using LLMs and specialized plugins to increase efficiency or quality or both within lines of all kinds of work.
- This applies to any firm, whether a web marketing or law firm or a nonprofit promoting Macular Degeneration solutions or a sawmill producing and selling wood flooring. Examples include generating marketing materials, proofreading web content and sales emails, making spreadsheets showing where sales teams can optimize results, scrutinizing office builder or IT firm contracts, tuning tax strategies, finding relevant files in a cloud, and countless more.
- The successful firms will determine whether or when each investment in an AI/LLM tool will pay off.
- Create project management systems that integrate all teams’ diverse specialists at just the right points in their lines of work in order to find holistic synergies among complex multi-part projects (the lines in the 2-dimensional model above).
- This pertains to knowledge work, like tax accounting or medical diagnostics, but it applies equally to knowledge-intensive projects at any firm, as in regional economic forecasting, production of marketing or training videos, and optimum arrangement of factory machines.
- Executives must stay minimally tuned to fields adjacent to their expertise and roles.
Now firms must allocate a substantial part of their R&D & training budgets to keeping abreast of AI/LLM tools. However, be aware that the AI/LLM field is in hyperdrive now, producing scads of startups and tools that seem amazing but are far from complete. Most will be coopted by big firms like Google and Microsoft. Often the few that are complete will not actually save you time and/or significantly improve results per unit of your time, when factoring their costs and your time to vet, set up, and tune.
What Happens After AGI?
Artificial General Intelligence (AGI) is the singularity event horizon nobody can see past. Elon Musk says it will be here between 3 and 6 years, and I read that the average of AI professionals’ estimates is 10 years. In the monumental 9/5/23 book, The Coming Wave, the author and co-founder of Deep Mind stresses that AI’s rapidly increasing capabilities matter more than do the future span of months during which consensus settles that AGI has in fact been achieved. When it is achieved, then, by definition, AI’s autonomous work will match or exceed all that consummate professionals do now, including innovative integrations. I strongly believe that when that time comes, the world will need a whole new social economy.
Will You Lose or Gain Income, and When?
It’s obvious that AI/LLM will change how we work, but the two overarching questions facing DISC and many kinds of firms and workers are:
- Whether or to what extent AI/LLM will reduce income within the next few years.
- How can one prevent lost income or even increase it by harnessing AI/LLM technology?
The projections below pertain to knowledge-based service firms, but the principles extend to most kinds of organizations. The timing may be off a year or two, but that would little change what you should start planning if not doing now.
Logically, technologies that let non-professionals conduct specialized work will erode consultancies’ aggregate revenues. Thus, we can expect accelerating declines in professional firms’ per capita billing over the next ~3 years. The main drive now in global AI/LLM investments is to reduce the time & cost of middle and upper management, as with Google’s Duet and the public firm C3.ai run by the founder of Siebel Systems, which was sold to Oracle for many billions. In web marketing this trend will be slowed by LLM-fueled increases in ROI, which pulls revenue from other budgets, but that only slows the trend. Within those averages, there will be an accelerating bifurcation between firms that integrate AI/LLM and those that don’t or do so poorly, resulting in fewer profitable firms within most industries. Small firms can be among the winners by deploying AI/LLM within main lines of work and by deftly integrating those lines. The “Q&A with 3 LLMs” below shows that such integration will remain a weakness of AI/LLM for at least 3 years, probably 5, possibly 10.
I expect that for about 3 years, large firms able to afford LLM technology will outcompete smaller firms, but then the tide will reverse as SAAS options enable small and medium-sized firms to integrate top-flight AI/LLM throughout their operations, and as consumers assert sovereignty “of the people, by the people, for the people,” to free themselves from corporate totalitarianism.
Q&A with 3 LLMs Regarding the Integration Imperative
The riveting conversation recorded in DISC’s public Google Doc, Lining Up AI/LLMs for Upper Management, supports this post’s central thesis summarized by Claud 2: “LLMs still lack the innate ability to autonomously integrate insights across datasets in creative and synergistic ways.” All three LLMs concur that such autonomous synergies won’t happen for 3 years, maybe 5, but almost certainly will in 10.
That “Lining Up” doc’s ~25 minute read reveals LLMs’ brilliance in explaining crucial distinctions; the ~4 minutes of bold text may convey the essence. The doc progresses as follows:
- Explains the problem of interdisciplinary synthesis in LLMs, in prompts 1 & 2;
- Further explains LLM’s inability to autonomously integrate diverse disciplines, in prompt 3;
- Resolves one crucial contradiction in the above, ultimately reinforcing the central thesis of this post, in prompt 4;
- Addresses the estimated years needed for AI/LLMs to essentially replace most middle and upper management, in prompt 5;
- Lists key vocabulary to help you walk in this new landscape.
I remain guardedly optimistic that the LLM’s support of my thesis is not pre-trained by their human creators to minimize threats to executives, in order to promote adoption and protect venture capital’s huge investments.
What About Your Search Marketing?
DISC has been using AI for years, especially in PPC and CRO. We started using AI/LLM’s earlier this year. We will be advising clients on setting up their systems to facilitate AI/LLM assistants for search marketing. That setup will help clients in all their knowledge work. Since its 1997 inception, DISC has waited for the ragged vanguard to settle into smooth proven results, an approach we bring to our ongoing evaluation of AI/LLM tools and processes. Avoiding dead-end paths, DISC will lead clients into a prosperous future, as we always have.
Finally and reluctantly, here are a few words about my relevant qualifications: Rob, Your Semi-Idiot, Semi-Savant.
See the follow-up post: GPTs for Upper Management. –>