Loading Article
Fetching the latest insights for you...
AI-Powered Strategic Consulting
Loading 0%
With over US$320 billion in 2025 already committed by Amazon, Microsoft, Alphabet on AI, where will future value from AI be created?
With over US$320 billion in 2025 already committed by Amazon, Microsoft, Alphabet and Meta on AI technologies and datacentre buildouts, AI seems unstoppable - is this sustainable or another bubble?
Longer term investors and strategists are seeking to understand how will AI deliver its long-promised productivity benefits, and where the next wave of AI powered value creation will happen.
A key development in 2024-25 has been the advent of commercially available AI Agents. By this, we mean Agents that can “reason”, breaking down problems into steps, planning for a solution and taking steps towards a given objective.
AI Agents typically comprise a "brain" which can be one of many LLMs (e.g. GPT 4, Grok, Gemini) to plan & reason, a memory (short or long term), and tools (to leverage external resources, systems, take actions like placing an order).
Examples of these include AutoGPT, ChatGPT Tools and Claude Agent. These tools can execute complex tasks by breaking them down into many sub-tasks, prepare a plan, execute tasks, and then consolidate results together. One of the early uses of agents are in customer service chatbots which can comprise multiple sub-agents to find customer information, generate responses, and filter responses against company standards before responding to a customer.
Over the last few years, most of the ‘killer’ AI applications we have seen to date were horizontal apps, providing general purpose capabilities that could be used across many industries (for example, ChatGPT and Copilot). However, we are starting to see a shift increased interest and investment in vertical apps, that is solutions that are industry specific, solving specific use cases very well and unlocking significant incremental value for users.
For example, coding is a leading use case for Agentic AI, particularly suited to Agentic AI because of the nature of coding being a text-based task with multi-step requirements, goal-oriented reasoning (e.g. solve to get this outcome), with output that can be deterministically validated through automated testing.
The strong product market fit in this vertical can be seen with the rapid growth of ARR for pure coding AI applications like Cursor that went to $500m ARR, and GitHub CoPilot with an estimated $300M+ ARR whilst Windsurf was acquired by Google for $2.4bn in a reverse acqui-hire deal announced on 11 July 2025.
For hints of other industry verticals and use cases where future agentic value may come from, we took a deep dive through Y Combinator's last three cohorts of investments (Autumn, Winter, Spring 2025) spanning 425 companies.
Surprisingly, we found that just under 80% of Y Combinator’s investments (by number of startups) were in horizontal AI, that is applying AI across a broad number of use cases. This may be understandable by the large TAMs that these investments can address across multiple verticals showing that there are still spaces to be claimed.
The remaining 20% of investee companies appeared to be highly specialised, vertical AI industry plays, with an outsized focus on healthcare, followed by fintech, insurance, legal and education. Examples include:
As customer and enterprise expectations increase, AI needs to deliver more value by taking action for the user. For example, in coding, this means tools that not only create code but also test and execute deployment into production. In other verticals, customer service agents that handle customer inquiries in both voice and chat format must resolve inquiries and integrate with operational systems to initiate refunds, closing cases out in real-time.
Despite Gen AI and Agentic AI feeling like a decade long journey, we are only at year two of a long multi-decade journey. Extending the “picks and shovels” thematic from the infrastructure layer, horizontal AI apps building generic “worker” skills and AI engineering platform tools have the potential to deliver deep financial returns from varied applications across multiple sectors as businesses build new tools and capabilities on top of this.
Challenging, industry specific problems are a hotspot for creating future value. Almost any existing SaaS business today could be done better with AI automation and adaptiveness. Targeted vertical specific solutions such as Heidi Health for example create deep value for customers from their proprietary system prompts, deep database of "learning data", patient histories and customised practitioner templates. Together these elements help to create deep value that is hard for new entrants to replicate. Founder teams with unique knowledge in verticals are one to watch and back.
Contact our team to discuss how we can help your organization implement these insights.