Shifting Techtonics
Insights on how to build innovation and ride the GenAI wave
Shifting Techtonics
Insights on how to build innovation and ride the GenAI wave
Although still only two years in to the generative AI revolution, it feels like so much has happened, and is yet to happen, such is the pace of change. Companies have much to consider in this dynamic landscape as GenAI middleware, GenAI applications and increasingly sophisticated use cases that use ‘agentic models’ build on strong foundations.
All of which elicits the key question: how do we innovate for the future? How best to formulate long-term value creation strategies, as this next wave of innovation builds momentum?
This was a dominant theme at Hg's ‘Shifting Techtonics’ Software Leadership Gathering in Lucerne in June 2024, where technology visionaries including 8VC’s Joe Lonsdale, Oracle’s global CIO, Jae Evans, and former VMware CEO, Raghu Raghuram, shared candid insights on where opportunities are arising.
Sometimes, the best way to think about the future is to look at the past and realise that throughout human history, certain behaviours (fear, greed) never change. And are history’s most important lessons.
Morgan Housel, partner at The Collaborative Fund and bestselling author of The Psychology of Money shared some fascinating insights on this during the leadership summit, while acclaimed three Michelin star chef Massimo Bottura, whose Osteria Francescana is a “laboratory of ideas”, gave his own unique insights on how to build innovation.
“If you want to build a multi-billion dollar company, you probably shouldn’t be doing something that people could have done 10 years ago, because it’s going be very, very tough.”
A marathon not a sprint
Back in the 1990s, strategy consulting began to grow as technology helped proliferate data. It effectively acted like rocket fuel and launched strategy consulting into a higher orbit, such that it now generates $40 billion to $50 billion in annual revenue. As the techtonic plates shift, it is not unreasonable to assume that other white-collar service industries will experience a similar degree of uplift. As such, when software CEOs look at all the hype coming out of the GenAI revolution, they would be best served to take a step back, and look at the opportunity set in terms of decades, not years nor months.
To build tomorrow’s technology unicorns, software leaders and investors need to consider what is possible now that was not possible five years ago.
AI services is expected to be an important area of disruption over the coming years. In the US alone, smaller services companies represent an estimated total addressable market of $4.5 trillion, a third of which could see their margins double or triple over the next five years.
Indeed, AI services has the potential to fully disrupt areas such as auto dealerships (where the customer service experience is ripe for innovation) and advertising, as businesses start to build digital ad campaigns using GenAI tools; thus avoid reliance on traditional advertising agencies.
One legacy service industry that is already being disrupted is wealth management. With $6 trillion of assets managed on its platform, Addepar is at the forefront of disruption. Its technology platform integrates vast data sets to deliver single consolidated portfolio insights to investors, in a way that previously wasn’t possible. By training its own LLM on this proprietary data, Addepar is already seeing the benefits of GenAI in helping clients make better decisions.
Rise of Agentic models
Now that generative AI’s foundations are in place, expect to see the rise of ‘Agentic’ models. These will not only contribute to further productivity gains for the software industry, but also automate workflows and begin to open a path to long-term revenue growth.
Many in Silicon Valley expect these architectures to be the next wave of innovation.
The emergence of an AI middleware layer will be critical to support Agentic models and broader GenAI application development.
This is already starting to gain traction.
AI middleware represents an important step in the technology’s evolution, as it will provide the scaffolding to make LLMs easier to manage: i.e. hallucination monitoring, LLM operations and orchestration etc.
The first wave of GenAI – and where the majority of use cases are currently focused – involved applying the technology to improve productivity. The second wave focused on customer operations efficiency. The third wave of GenAI will involve companies building it into their core software products, which will also help justify rising software prices.
Strong AI foundations
To do this, however, will require a committed shift in thinking and willingness by software CEOs to act quickly…even if this means creating a dedicated skunkworks team. Indeed, not enough attention is yet being applied to the underlying technology, or ‘platform shift’.
Looking at the data, it suggests that there is a real, immediate efficiency opportunity to companies, worth 2 to 3% of margin improvement.
However, the level of innovation – and efficiency gains - is a direct corollary to how powerful the use cases are. Businesses beginning their GenAI journey should first start with productivity and efficiency gains, before moving on to revenue growing initiatives.
When longer-term monetisation opportunities do arise – as could be the case with Agentic models – it will be incumbent upon software firms to have strong AI foundations and change leadership, to drive meaningful value creation.
For now, the industry is still waiting for that first killer app to emerge, akin to when Uber first appeared in the App Store in May 2010.
“Historically, we always overestimate the speed of change, but we underestimate the scale of change.”
- David Toms, Hg
“The commonality is the use case that even a middle schooler could understand, appreciate and get excited about. When these things happen enormous amounts of money pour in and people start using it without regard to things like ROI…the greatest technologies always happen when people ignore ROI.”
“The greatest technologies always happen when people ignore ROI.”
An engineer-led culture
Software CEOs need to consider their approach to build and foster innovation within their organisations, if they intend to stay on the bleeding edge of AI disruption. Getting things right today could lead to tremendous earnings growth in the years and decades ahead; not in 12 months’ time.
For start-up CEOs, this should not mean that they try to imitate what others are doing. Instead, they should take a first principles approach. The bar to compete against large established SaaS businesses is, arguably, too high. That being said, for software businesses of all sizes, the best way to build innovation is to promote an engineering-led culture, where developers have the latitude to think big, fail, and think again about innovative product development.
What is needed to build the right technology that could in turn help other businesses get stronger? Hiring exceptional talent is one aspect. But another is to avoid creating too much of a sales dynamic. This can quickly lead to mission creep in respect to product development and is typical failure mode for innovation culture.
When it comes to striking a balance between breakthrough versus iterative technologies, therefore, organisations ought to strive to keep the innovators separate from the adults. Creating the equivalent of a factory floor is necessary for any breakout product to achieve scale but the laboratory is where the magic happens. Both are needed.
Software technology companies who are pioneering the GenAI revolution are already seeing the fruits of their innovation labours. By successfully using GenAI in its code development tools, Salesforce is saving up to 20,000 engineering hours a month. The next step is how to deploy GenAI in such a way that its customers, too, can reduce their workload, without having to build their own LLM. Larger, mature companies like Salesforce can choose to focus on revenue growth activities. Start-ups, by contrast, are achieving the most success using GenAI tools to drive productivity and efficiency gains.
In short, companies need to know where they are in their AI journey.
As was mentioned during this year’s SLG, the biggest mistake with product innovation is thinking, ‘Here’s how the world should work’ rather than, ‘This is how the world does work’.
A fundamental shift in the application stack
Looking ahead to the next wave of GenAI, and what this might look like as AI middleware builds on the technology’s foundations, there is a sense that the application stack will fundamentally change. What this means is that GenAI applications will become intrinsically tied to LLMs. Just as APIs rose to prominence in the 2000s, autonomous agents are expected to do likewise in the coming years.
As LLMs are seemingly being released on a monthly basis, it offers a huge potential for the orchestration of conversational AI tools across a wide range of B2B workflow processes. Indeed, such is the pace and proliferation of LLM, some software leaders hold the view that they will become commoditised. And that the easiest place for SaaS companies to invest, going forward, will be in vertical AI solutions such as application software, more so than horizontal AI, which is not regarded as a sustainable venture model.
The rationale for this is based on the fundamental advantage of proprietary data sets.
Companies that collect and store private data will have, in effect, a walled garden that lies beyond the perimeter of generic LLMs. As the next wave of GenAI builds on top of AI middleware, this will allow companies to begin to build their own domain-specific large language models, at potentially less cost. GenAI applications that operate on top of these bespoke LLMs could unlock an array of new customer and product insights. Vertical AI is a real opportunity.
Enterprise AI and the need for data sovereignty
Data privacy and data sovereignty are salient issues today, as cloud providers consider the best ways to augment the client experience while also mitigating risk. In a fast changing environment, cloud providers need to provide a highly secure, robust, flexible platform experience that offers the controls that enterprise businesses need. When CEOs put their trust in the cloud, they want to know who has access to their data. How is it shared? In fact, trust is of such high importance that it is holding back the rate of public cloud adoption.
Moreover, there is a desire for choice, whether in a single or multi-cloud arrangement. Some enterprises just want the basic infrastructure on which to build APIs, while others want to leverage SaaS solutions provided on the cloud.
Data sovereignty and flexibility are combining today such that enterprises are looking more at private cloud options. Not only will more infrastructure be needed to meet this demand – requiring cloud providers to think carefully about their point of differentiation. But as middleware develops, there will be an emphasis on the integration of in-depth multi-layer security to protect client data.
Trust and verify
Companies like VMware are seeing tremendous interest in their private cloud offering. As the GenAI revolution began in earnest, CEO Raghu Raghuram began thinking, ‘How can we begin to use GPUs instead of CPUs?’
The private cloud is benefiting some of VMware’s clients who want their data to remain within their respective jurisdictions. But more than that, it is allowing them to verify answers/outputs from LLMs trained solely on their data, which isn’t possible when using public data-trained LLMs. Trust and verify: two key factors as private AI cloud solutions gain favour.
To put this into a wider context, today the global GPUaaS market is worth approximately $4.31 billion. By 2032, it is forecast to be worth $49.8 billion.
Final takeaway
Software CEOs face a lot of uncertainty as the GenAI revolution unfolds and impacts global business. Rather than get lost in the noise, it will require tuning in to a signal frequency that fits their business model. With the right commitment to change leadership, an engineering-led culture, and understanding where they are on their AI journey, software CEOs will be at an advantage. This is the time to think and embrace innovation.
Photos courtesy of Laurin Grether, Grether Photography GmbH