TL: DR — Our predictions
- Robust IPO Market: With a strong economy and a venture-friendly Presidential administration, the IPO market is set to thrive in 2025, evidenced by accelerating deal values and significant capital raises.
- AI Agents will become a reality: AI agents will overshadow traditional software applications built on LLMs, with barriers to adoption lowering to make way for agentic workflows.
- Service-as-Software Models: AI-driven service-based businesses will shift from hourly billables to subscription models as they streamline client service delivery.
- Enterprise Build vs. Buy Shift: Enterprises will increasingly choose to build custom AI-integrated solutions over buying off-the-shelf software driven by AI-native low/no-code tooling.
- AI-Native Solutions in New Verticals: Founders will develop AI-native applications for new industries, eventually creating niche solutions for specific workflows and end-users.
- Rise of Digital AI Workers: Digital AI workers, such as those provided by companies like 11x for sales and go-to-market roles, will see significant growth, outpacing other generative AI applications.
- Re-entering the Infrastructure Layer: Early-stage startups will have opportunities to innovate at the infrastructure layer of AI development, focusing on reasoning and inference improvements over sheer data volume.
Ardent is excited about 2025. With the strength of the current economy and an incoming Presidential administration that supports the venture ecosystem — including plans to scale back regulation and bolster tech sectors like AI and crypto — we predict an improved IPO market next year. We've observed deal values accelerating towards the end of 2024, with capital raised reaching $67.9 billion through October. If ServiceTitan's initial successful IPO is any indication, investors will have reason to be optimistic about an expanding pipeline of exits for their best portfolio companies. 2024 had 30 billion-dollar exits, up from 18 in the year before, underscoring the growing appetite for strategic M&A deals, signaling renewed activity and another way to return capital to investors and LPs.
With this environment as the backdrop, early-stage venture will continue to focus on generative AI, which has been a bright spark in a market still recovering from the highs of free-flowing capital in 2021. The ongoing shakeout will deepen, with only the best funds able to raise capital from LPs and slow-growing and indefensible startups going out of business. Despite this, generative AI has made it possible for companies to grow at a pace we have not seen before across our venture careers. AI founders are achieving more with fewer resources, leveraging LLMs to streamline functions such as marketing and sales while maintaining capital efficiency. These next-generation startups — often fewer than 10 people — are scaling to tens of millions in revenue, and eventually, we may even see our first one-person unicorn.
So what do we think will happen next year as generative AI continues to eat the world? Here are our five predictions.
AI agents will emerge in both enterprise and consumer contexts as a viable option instead of traditional software applications atop LLMs. We predict that some of the barriers previously hindering this area will be resolved early in the year, making it easier to research, build, sell, and adopt agentic workflows.
Read more: AI Agents Show Huge Promise, But Face Technical Barriers to Wide Adoption
AI-driven service-based businesses will continue to emerge and scale quickly into attractive, venture-backed opportunities. Some may begin to shift their business models from hourly billables to subscriptions as they reduce the time needed to produce for clients.
Read more: You've heard of software as a service; enter service-as-software enabled by generative AI.
More enterprises will opt to "build" software as the cost to develop decreases, driven by AI-native low/no-code tools. Given the shift away from buying software solutions off the shelf, existing software vendors will need to embrace AI to become more malleable.
Read more: Build vs. Buy: The coming revolution in enterprise software
Founders will build AI-native vertical solutions in new industries, starting with broad-based applications and eventually niching down into specific workflows and end-users. In more mature industries that have seen a lot of activity, startups will offer hyper-specific solutions to differentiate themselves.
Read more: One Year Later: A Refreshed View of the AI-Native Vertical Software Landscape
Related but separate from our prediction about AI agents, we also believe that digital AI workers will be a hotspot of activity within GenAI applications. Companies like 11x, which offers digital workers for Sales, RevOps, and Go-to-Market jobs will continue to grow faster than other areas of generative AI applications as companies "hire" their workers.
Lastly, startups will re-enter the infrastructure layer
A year ago, the conventional wisdom was that competing with models from hyperscalers like OpenAI and Anthropic required vast resources. The assumption was that enhancing the performance of LLMs necessitated training on more data than the rumored 13 trillion tokens used for GPT-4, leaving little room for early-stage startups in this space. We also believed that there was no room for new entrants in this arena.
Since then, the landscape has begun to evolve. Many believe the approach of pre-training LLMs with additional data is reaching its limits as most existing and public data has been utilized, and synthetic data has not yet (to our knowledge) proven effective in filling this void.
We predict two implications for early-stage companies. First, there will be increased stability for startups to build upon the latest models without fear of a better, shinier version disrupting their progress. Second, this shift opens the door for improvements in model performance that don't rely on vast data resources. Not only may hyperscalers focus on enhancing the reasoning capabilities of existing models, but startups may also venture into this arena as well. The reasoning and inference layer could become the new frontier for differentiating models, requiring not vast data but rather verticalized, hyperspecific datasets to remain competitive. We are excited to see how this unfolds.
Thank You for a Great 2024
As we look to 2025, we at Ardent are ready to embrace the opportunities and challenges the new year will bring — whether it's the rise of AI agents, or the transformation of service businesses, it is exciting to see these things become realities and solve real use-cases where they can add value. We are deeply grateful to those who have joined us on this journey, sharing insights, engaging in dialogue, and contributing to the vibrant discourse around these trends. Together, we look forward to pushing the boundaries of innovation and progress in the new year!
