Over the past year of generative AI breakthroughs, most people have been talking about use cases that are not that much different from what machine learning offers automation, productivity, recruiting, accelerated methods of classification, prediction, and natural language processing.

What they are not talking about is how generative AI will play an outsized role in strategy development – both at the corporate level as well as at the functional, business unit, product, and brand levels. And how the all-important “first mover advantage” will deliver superior returns to those who act now.

The conundrum most companies are faced with is that they feel they can’t wait and expect to catch up later. However, the high costs of generative AI software development and the intense war for AI talent are raising new barriers to competitive entry for even the fastest of fast followers.

The high costs of generative AI software development and the intense war for AI talent are raising new barriers to competitive entry for even the fastest of fast followers.

Fast-moving innovation demands fast strategic choices – and the smart use of generative AI helps to move fast. The good news is that it is straightforward to fold generative AI capabilities into an existing strategy process and frameworks for using enterprise software in strategy and operations.

The Dearth of Useful Use Cases
By this time, almost everyone has tried ChatGPT, Microsoft Copilot, or other generative AI tools in one form or another. Most busy people have found these tools to be useful at some point to get a specific task automated. They have given a shot at some task “in the moment” and, when it didn’t live up to the hype, they moved on.

For example, one executive needed a quick immediate summary of trends in their industry. The AI-generated results were similar to the output from an incomplete search engine – a series of links to publications rather than the desired summary of the publications’ content with something even more intelligent and useful on top.

A new AI strategy will result in new data to be used as input to the next cycle of strategy formation.

I hear this complaint time and time again. A group of middle managers, over drinks, compared their first attempts to get something out of this innovation. The one who gets the most laughs used it to generate the top ten lists for travel. Not exactly earth-shattering stuff.

As many have already observed, there are tiny, incremental improvements in productivity that each of us can eke out of these first-generation tools. Nudges to help us think more broadly, use more comprehensive spell-checking, translation, and, yes, search. However, beyond the immediate gains experienced by die-hard software coders and brand marketing professionals, there is a dearth of useful use cases out there.

This lack of great use cases may lead to cynicism, but what we witness are just the same adoption dynamics as in previous waves of software innovation (think about the first spreadsheets, word processors, and graphics tools). Over time, most people will experiment with these generative AI tools in the course of their daily work, until better versions, more applied versions, come along. And they learn in the process.

Companies that are Bypassing the Hype Curve
There are a few market leaders that have already gone past this phase. They are going beyond “hopping on the hype curve”.

Adobe is a prime example. Rather than wait for newcomers Midjourney or Stable Diffusion to disrupt their hold on the creativity marketplace, Adobe created and rolled out their version of generative AI for visuals, called Firefly, to much acclaim. Firefly not only works simply and easily within designers’ workflows, but it also deftly avoids any potential legal issues by not being trained on the unauthorized intellectual property of others.

Another example is Accenture, which announced that it will invest up to $3B to train an army of its employees to be AI-focused, both in order to use generative AI more in its client work, as well as support those customers in their use of the technology. All on the heels of similar press releases from PwC, EY, Bain, Deloitte, IBM, BCG, and McKinsey, who has already rolled out a ChatGPT-like bot called Lilli.

To use GenAI effectively, it is important to apply your existing know-how in identifying and mitigating biases and errors before your customers do.

And big tech has taken note, of course, as it faces potential disruption.

Alphabet is on ‘code red’ to protect its core search engine business, trying to offer all the excitement of the new technology to users that opt-in, without taking the unnecessary risk of upsetting the status quo. Through this “labs approach”, including partnering with new players such as Anthropic, and systemically rolling out Gemini, Alphabet plays to its strengths as a data giant, gathering a flywheel of user information to improve its foundational models and applications.

Microsoft, on the other hand, with OpenAI and ChatGPT in its back pocket, sees only an upside in dominating the narrative of being an innovator again. It is using its massive distribution advantage and installed base of enterprise customers to keep pushing the boundaries of when, where, and how a person develops the habit of relying on one of its software “co-pilots” or agents to get work done.

Even Apple is rethinking its strategy, announcing the release of its model, Ajax, and its chatbot (and Siri replacement) codenamed Apple GPT, in conjunction with its new iPhone 16 line-up.

What are the implications for the vast majority of companies who are faced with the above-mentioned conundrum?

Opportunities Abound…
In times of uncertainty and disruptive change, it makes sense to go back to basics. Corporate strategy means making clear choices about what to do and what not to do as a business. Corporate strategy looks at capabilities that competitors don‘t have or would struggle to have and, based on these insights, targets markets and shapes its products and services accordingly. New technology like generative AI does not change the fundamental questions but has a lot to do with how to answer them. It is important to recognize that generative AI tools are simply the latest innovation in a company’s enterprise software “stack”. And, just as with their existing software, companies can take advantage of these new generative AI tools to refresh and stress-test their corporate strategy.

Take for example brainstorming a key practice when it comes to developing novel strategic ideas. GenAI is radically changing the game here. Not unlike scientific researchers in the life sciences, companies can set up a literature-based discovery (LBD) process to look for hypotheses, connections, or ideas that humans may have missed. They can expand their training datasets to include every idea they ever considered, those chosen and those rejected, along with the rationale. It is possible to create a proprietary chatbot trained on this internal dataset and query away. Companies can apply tests that allow them to challenge assumptions. They can broaden their constraints with externally acquired datasets from a diverse range of sources. Experiment with “how to win” dimensions such as the degree of autonomy, ranging from fully autonomous to human-in-the-loop. Add more data. Iterate, iterate, iterate. These are leapfrog opportunities for a creative strategy process.

Capitalizing on these opportunities requires more than just investments in technology. It requires talent that is not only AI-savvy but also has the domain expertise and experience to assess if these words and numbers that are generated by the new tools make sense.

Existing analytical tools are about to change dramatically, too, as they present an interface layer that is intelligent and driven by underlying algorithms. Companies like Microsoft, Alphabet, Salesforce, and Adobe are offering their next generation of office productivity software – spreadsheets, data mining, visualization, observability, design tools, CRMs, and more – accelerating turnaround times on each task and query. Companies will need a team of analysts able to maximize the value of these new applications, working alongside traditional strategists. If they manage to do this well, they will gain a competitive advantage.

Lastly, an exciting feature of corporate strategy in a time of generative AI is that properly designed and executed, a new AI strategy will result in new data to be used as input to the next cycle of strategy formation. In other words, we may see the start of a more dynamic system of strategy-in-use.

Capitalizing on Internal Data
Much has been written about the ‘rough and ready’ nature of the outputs of Generative AI. They have even coined the phrase ‘hallucinate’ to describe the errors introduced by the software. It is a classic case of GIGO (garbage in, garbage out). You need to know the source datasets used for training the algorithms and for building the LLMs (large language models) to determine the quality of the output. To use GenAI effectively, it is important to apply your existing know-how in identifying and mitigating biases and errors before your customers do; you must pay special attention to inputs, models, and algorithms.

Given the complexity and ‘black box’ nature of the new tools, strategists, and executives need the support of AI specialists capable of explaining the underlying rationale for which AI generated option rises to the top (which requires expert knowledge of the input datasets and tools used), and what the tradeoffs are by not going with the second or third placed option.

Equally important is the issue of relevance, or specificity. Depending on a company’s strategic goals, there will be preferences and priorities as to which datasets should be used. Companies may even want to include their proprietary datasets, especially custom data that grow and refresh over time.

Models get stronger if they are fed a company’s entire history of internal documentation, across functions, along with financials, communications, product performance, etc. No longer are there data-related limits to what can be incorporated into the models. As Adrian Cockcroft of OrionX.net stated, “You’re fine-tuning the model to be somebody that understands your company.”

There are many different ways to marry external AI tools with internal data so that ChatGPT can then generate responses customized to a company’s unique circumstances. One option is to fine-tune the GenAI engine against proprietary datasets, which requires in-house expertise in API access and in the use of software libraries such as TensorFlow or PyTorch to do the heavy lifting of the training process. Or you may choose to refine your prompt engineering methods against your database, and then use those custom-tailored prompts to query the external engine. However, both options present trade-offs in terms of accuracy and expense.

A Back-to-Basics Game Plan
To take advantage of these opportunities, it is important to candidly assess your capabilities and make a ‘buy’ versus ‘build’ decision. Large-scale digital natives such as Amazon or Microsoft enjoy a very different talent and capability situation as industrial legacy players. Depending on the various contexts, companies will have to choose their distinctive AI strategy. How much should they leverage external tools and data, and how much can or should they develop inside to best capitalize on their internal data sets? Is it better to create an in-house LLM or fine-tune an open-source LLM on their data?

The answer is it depends, as both approaches have their pros and cons: In-house LLMs offer the advantages of full data ownership, more control, and lower cost of inference (the cost of asking the LLM to generate a response), but deliver lower performance along with greater risk and cost of safety.

APIs (application programming interfaces that connect with external solutions) are the mirror-image; they offer the advantages of higher performance along with the ability to offload risk and cost of safety, but require the sharing of data, resulting in less control, and a higher cost of inference.

While the trade-offs are clear, the costs and complexities are significant. If your company is among the fortunate few that have the abilities and resources, you may have already built or are building a contender. Most companies are not that lucky; and just like with any new category of enterprise software, they need to choose between experimenting and the available commercial options.

One executive described how they use multiple suppliers, how each is slightly different, and how each generates content that is beneficial for different use cases. By embracing this diversity, they are building in-house proprietary expertise in how to use each piece of the puzzle to its maximum potential.

If these issues sound familiar, it is because they are. We have seen the same dynamic in cloud-computing when companies debate about proprietary versus hosted solutions.

A New Set of Capabilities
It is evident that each step of the corporate strategy process – from picking markets and products to deploying capital and capabilities – can benefit from incorporating generative AI tools. However, capitalizing on these opportunities requires more than just investments in technology. It requires talent that is not only AI-savvy but also has the domain expertise and experience to assess if these words and numbers that are generated by the new tools make sense.

In this context, it is important to revisit the assessment of a company’s capabilities. What are the pre-existing and/or nascent skills and resources critical to the use of generative AI? What must be developed or sourced from outside experts? Be brutally honest as to what your company can and cannot do, and what it may take to address any gaps. So yes, corporate strategy changes in a time of generative AI. What the internal team of strategists does changes. What external consultants do change? The capabilities needed to succeed change, and so does the need for collaboration in the AI-related ecosystem. What probably won’t change are the investments in strategy. Like the waves of automation and robotics in the past, the advent of generative AI itself does not threaten you or your company. It is the sophisticated use of generative AI by your competitors that is the threat.