A Year In Review Of IBM’s Ambitious AI Strategy

By Patrick Moorhead - February 22, 2024

We will look back on 2023 as a watershed year for AI. Starting in late 2022, when OpenAI introduced ChatGPT to the world, there was relentless hype for generative AI in all its forms. It was a hype train that many technology companies boarded enthusiastically, as there were board of directors’ mandates to the leadership teams of major enterprises to “make something happen.” But there was real substance there, too, as companies from Adobe to Zoom demonstrated by introducing AI products that genuinely moved the needle in their respective markets.

IBM is firmly in the “real substance” camp, which makes sense when you understand just how large its AI research and development program is, how long that program has been running, its enterprise-first focus and how quickly its AI products have moved into GA. For example, see my colleague Paul Smith-Goodson’s analyses of IBM’s work in foundational models (February 2023) and its creation of a giant cloud-based supercomputer to train AI models (November 2023). IBM is now adding to that background with big launches for the watsonx platform and a slew of Granite AI models. It is also helping thousands of clients across multiple industries operationalize AI for human resources, customer experience, application modernization and more. This is a long way from jumping on the AI bandwagon—this is the real deal.

Full disclosure: IBM is a client of Moor Insights & Strategy, as are all of its major competitors and many other companies active in AI, but this article reflects my independent viewpoint as an analyst.

Building On Years Of AI Momentum

Some of the most famous AI milestones in IBM’s history came from famous victories in games. Its AI supercomputer Deep Blue beat Garry Kasparov in a six-game chess match in 1997, and then the Watson system beat Ken Jennings on Jeopardy in 2011. The company wasn’t building those computers just for show; it quickly put them to use in areas ranging from cancer treatment to retail optimization. Five years ago, Rob Thomas—then the head of IBM Analytics, now IBM’s chief commercial officer—was advocating tirelessly for companies to try many smaller and larger experiments to put AI to practical use in serving customers and finding better ways to solve business problems. My colleagues and I covered these enterprise AI efforts in detail (for example in 2016, 2017 and 2019).

Those of us studying AI back then saw what many companies have embraced more recently: that AI will be transformative for many, many areas of work. And although GAI has dominated the broader discussion for the past 15 months, it’s no mistake that technology leaders including IBM, AWS and Google continue to make advances in other facets of AI as well.

Capitalizing On Watsonx

For IBM, a lot of this work is reaching customers today via the watsonx AI and data platform, launched in May 2023, which uses not only GAI but machine learning, deep learning and foundation models. Within the platform, watsonx.ai is an AI studio that supports all phases of AI development and deployment; watsonx.data is an open data lakehouse optimized for AI; and watsonx.governance takes advantage of IBM’s deep roots in data governance to ensure trust, accountability and transparency for AI processes and projects. Built on Red Hat OpenShift, the platform is designed to be flexible enough to accommodate widely different functions—anything from cutting-edge data science to the nitty-gritty of optimizing financial portfolios, marketing campaigns or supply chains. It can also automate specific tasks from software coding to compliance reporting, in many cases reducing costs and cycle times by more than half.

Forbes Daily: Get our best stories, exclusive reporting and essential analysis of the day’s news in your inbox every weekday.

Since launch, IBM has kept improving the platform. In some cases, this has been with technical items such as new vector database features. But it has gone much bigger than that with the release of its Granite series of GAI models. These models are trained on specially vetted and curated datasets from five key domains—internet, academic, code, legal and finance—for use in enterprise apps and workflows. Customers can pair the models with their own data to build customized models large or small. To extend the utility of this, IBM has also enabled the use of third-party models, for example Meta’s Llama-2 or models from specialist AI providers such as Hugging Face.

Using the right combination of models from IBM or third parties with your own data can be a powerful mechanism to create new insights about your market, make informed decisions and respond to customers in ways uniquely suited to your company’s needs. IBM’s benchmarks show that the tailored models increase accuracy, relevance and price-for-performance—so customers get the best of everything.

The AI models and the broader watsonx platform fit into a bigger GAI stack, as shown in the image above. The stack is strengthened by large pools of expertise within IBM, for example in OpenShift, data services (where the company has long been a power) and integrations; it is further supported by IBM Consulting and IBM’s extensive network of system integrators and other partners. Just since May 2023, IBM has formed new partnerships with SAP, AWS, Adobe, Microsoft, Salesforce and others to increase enterprise adoption of enterprise AI.

Without Strong Trust And Governance, Enterprise AI Is A Non-Starter

IBM is sure enough of what it’s doing that it provides full indemnification for its models, which is a huge selling point for enterprise customers. The models’ training data is filtered to mitigate concerns regarding privacy, bias and objectionable material while upholding high standards for AI transparency and explainability. This is right in line with IBM’s robust overall stance on AI governance.

The company’s approach to this reminds me of what two of its corporate partners, Adobe and Salesforce, are doing to ensure the responsible use of AI at enterprise scale. Adobe is offering IP indemnity for customers that use its Firefly AI art generator. Like IBM, Adobe can do this because it has so carefully vetted the library that supports this GAI functionality. Meanwhile, Salesforce has built a “trust layer” into its new Einstein 1 AI platform (which I analyzed in depth here). This allows customers to use LLMs from Salesforce or third parties without putting their own sensitive enterprise data at risk—while also closely checking the results to eliminate toxicity or AI hallucinations. All three companies have spent decades earning their stripes as upstanding corporate citizens, and kudos to them for doing the hard work to create a high level of transparency and assurance for their enterprise AI customers.

Launching The AI Alliance

Also in the vein of advancing trusted AI, IBM has cofounded the AI Alliance alongside Meta. The group was formed to support open innovation across the AI landscape to accelerate progress, improve safety and promote security and trust in AI. It has more than 70 members worldwide, including universities, agencies and nonprofits alongside startups and major tech companies such as AMD, Dell, Intel and Oracle. The AI Alliance intends to bring together a critical mass of compute, data, tools and talent to increase open-source innovation across AI software, models and tools.

It’s interesting to note that the AI Alliance does not include some of the major AI players, most notably Amazon, Google, Microsoft, Nvidia and OpenAI. Those companies may not be as eager as IBM to support the AI Alliance’s specific goals for open collaboration, governance guardrails and so on, or they may have other reasons of their own for not joining just yet. It remains to be seen whether all the leading firms in the AI sector will come together on these matters or end up supporting disparate approaches.

Real-World Focus Areas And Emerging Trends For IBM AI

Meanwhile, IBM is busy implementing and scaling AI across many real-world settings. The company has been baking AI into its own operations for years—a practice that has only accelerated in this new era of GAI. More than that, it’s the only major technology vendor that also has a full-service consulting arm, so it’s able to observe and guide what many other companies are doing in AI as well.

IBM Consulting has identified five big areas for GAI implementation:

  • New business models — Many companies are working to create new revenue streams via digital products based on LLMs.
  • Reimagined customer engagement — The opportunities are endless to use GAI for tailoring and streamlining customer interactions, from better-targeted marketing to quicker resolution of complaints.
  • AI-powered decision making — GAI can help business experts make better decisions within their specialized domains, whether that means managing an investment portfolio or more accurately detecting banking fraud.
  • Extreme process digitization — The automated compliance reporting mentioned earlier is one example among many where thoughtfully implemented GAI processes can reduce cycle times by 60% to 90%. Applying improvements like that to even routine functions such as invoice processing or contract management can drive down operational costs, often while also increasing accuracy.
  • Content and code creation — This is where a lot of the public’s attention about GAI has gone, especially given the prevalence of automatically generated text and images among consumers. The challenge for enterprise AI is to do this with precision, whether in critical business processes or new products. Among other examples, IBM has worked with partners to create auto-generated spoken AI commentary for sporting events such as The Masters golf tournament and the U.S. Open tennis tournament.

The examples from The Masters and the U.S. Open are good for reminding us that the applications of GAI, even for something as “simple” as content creation, will go far beyond what we’ve seen in the early days of people asking ChatGPT to write a Shakespearian sonnet about the Kansas City Chiefs. In fact, through a partnership with Adobe, IBM is training and tuning a “brand brain LLM” that will know your brand and content guidelines—and how they should apply to individual pieces of content your company produces.

These functions also benefit immediately from using the smaller, more tailored models mentioned above. IBM has been helping customers move rapidly from pilot programs to large-scale implementations, syncing models across multi-hybrid environments—which is a must, especially because many of the smaller AI models live on the network edge and in private clouds. (That said, IBM also needs to keep a sharp eye on what the big public cloud providers are doing. In particular, it can count on AWS infusing generative AI into IaaS and PaaS at scale.)

IBM is also drawing on its expertise in data management to make all of this work. According to Glenn Finch, a managing partner in IBM Consulting who is leading projects like the ones mentioned here, “These data pipelines are radically different than what we’ve seen for the last two and a half decades in traditional data platforms.” Based on its history, if any company can master these new data pipelines, it’s IBM.

What Lies Ahead In AI, For IBM And Others

After listening to IBM CEO Arvind Krishna’s keynote at the Think 2023 conference in June, I wrote, “IBM wants to be your go-to provider for enterprise generative AI. The company offers the necessary tools, data layer, governance and fabric to build the needed workflows, but its speed to market remains a critical question.”

Color me impressed with what Big Blue has done since then. I’ve been tracking Big Blue for decades, and I can tell you that the level of energy, excitement and execution I’m seeing around its AI efforts—from IBM and its customers alike—is something special. And they’re not tackling only siloed problems. In a recent briefing, leaders at IBM Consulting made the case that GAI could lead to major, across-the-board improvements in business functions including talent acquisition/management, customer service and software development. We’re talking about a 40% improvement in overall HR productivity, more than 90% of customer inquiries handled by AI assistants and—this one really staggers me—up to 60% of software development content automatically generated by AI.

Folks, this is what I mean when I say that AI will fundamentally change the way we do business. It won’t just be sporadic or limited to a few functional areas; it will rewrite whole industries and entire career categories. IBM has plenty of challenges ahead of it to fulfill its ambitions in AI, not least the competition it faces from a Who’s Who of major technology companies with even bigger resources than it has. But given its long history in AI, its specific domains of expertise and the depth and breadth of its approach to scaling AI to solve real-world problems, I think IBM is as well-situated as any company to make the most of what this new era will bring.

I’ll end this analysis by saying, and I don’t use these words lightly, that we are looking at a new IBM. The refocus on technology, the pace of change, service and product time-to-market and the partnerships that were formed across system integrators, ISVs and even the largest cloud service provers is a new look. A better look. IBM mistakenly gets overlooked given its important but very different past. If you haven’t looked at IBM lately, whether you’re an enterprise, partner or even investor, I’d recommend a fresh look.

Patrick Moorhead

Patrick founded the firm based on his real-world world technology experiences with the understanding of what he wasn’t getting from analysts and consultants. Ten years later, Patrick is ranked #1 among technology industry analysts in terms of “power” (ARInsights)  in “press citations” (Apollo Research). Moorhead is a contributor at Forbes and frequently appears on CNBC. He is a broad-based analyst covering a wide variety of topics including the cloud, enterprise SaaS, collaboration, client computing, and semiconductors. He has 30 years of experience including 15 years of executive experience at high tech companies (NCR, AT&T, Compaq, now HP, and AMD) leading strategy, product management, product marketing, and corporate marketing, including three industry board appointments.