Predictions for 2025 in Generative AI

17 educated guesses about what might happen this year

Happy New Year! I hope everyone had a nice holiday break. I expect 2025 to be a big year for GenAI where we will see the technology work itself further into the mainstream. I’m hoping we will see lots of examples this year of companies using GenAI to achieve real financial success and returns for investors.

With this post, I will attempt to forecast what’s ahead. With these predictions I’m trying to be as specific as possible, so that in a year, we can look back and see what I got right. I encourage you to weigh in on these predictions, challenge them, and add your own in the comments.

1. No release of GPT-5 or equivalent, but smarter thinking models emerge

• The era of "bigger is better" in AI models seems to be over. I predict that nobody will release a breakthrough model built by training on a larger dataset or number of nodes.

• But, we’ll see new models that can solve harder problems by "thinking" longer such as the o3 model that was announced before the holidays. I predict that there will be a new o4 model, capable of tackling increasingly complex reasoning tasks. Investors should be looking for ways that these “thinking” models can drive new use cases for businesses, an exciting topic to explore this year.

• The new thinking models will lead to AI solving at least 1 problem in math or physics that humans had previously not been able to crack. This will be a big deal, but it will be a highly theoretical problem with limited practical value.

2. Agents take center stage

By 2025, GenAI agents will become able to handle tasks like sending emails, modifying calendars, and placing phone calls through widely available tools like ChatGPT. B2B agent applications will also proliferate, creating opportunities for businesses to build bots that can take specific actions on behalf of users, but adoption will be slower than in consumer applications.

3. Self-driving cars go mainstream

• Self-driving taxis: Waymo will successfully expand to at least two more cities and double its ride count to 200,000 per week by year-end

• Cultural acceptance: Mainstream awareness of self-driving taxis will grow, and I predict there will be a TV episode where one plays a central plot role

• Self-driving trucks: While they won’t dominate highways, short-haul routes will see meaningful adoption. At least one company will exit pilot phase, operating at significant volumes for short-haul trucking within a state.

As I’ve written about, the autonomous vehicle market is finally becoming real, and it will start to have real effects. No jeansmobile in 2025, though. Maybe 2026?

4. AIO (SEO for GenAI) emerges as an important topic

As generative AI systems become the starting point for search, businesses will need to optimize for these systems much like they do for Google. You see more on this here: B2C and B2B. I expect:

• By year-end, most large digital agencies will offer AIO (AI Optimization) services.

• Paid content will be part of the GenAI experience, with at least one player—Perplexity, OpenAI, or Claude—integrating ads into at least the free version of their products

Companies (B2B and B2C) will need to carefully monitor their AIO position and invest to improve it.

5. Token costs continue to decline

Despite concerns that LLMs are waiting for the right moment to raise prices, I predict GenAI token costs will drop by at least 10% by the end of 2025. For instance, an o3 equivalent will cost less than an o1 model today. This trend will lower barriers to adoption and make it easier for companies to add advanced GenAI features.

6. Video models cross new frontiers

GenAI video technology will continue to improve, and by the end of 2025, the first commercial made entirely using generative AI will air on TV. However, full TV episodes or movies created solely with GenAI will remain out of reach for now.

7. GenAI-Powered Call Center Bots Expand

Generative AI will become a staple in call centers, reducing human-handled call volumes by 5-10 percent. Companies embracing this technology will see both significant cost savings and improvements in customer satisfaction.

8. Sales enablement bots flood inboxes

AI-powered sales tools will reach critical mass. Firms will use tools like Agentforce to enable their salespeople to be more productive, sending more targeted, higher quality, and more frequent communications. I predict I’ll personally receive 10+ emails or LinkedIn messages clearly generated by bots, pitching B2B solutions. This proliferation could lead to a one-time increase in sales productivity. Investors will update their commercial excellence playbooks to incorporate these new technologies

9. GenAI robots enter warehouses, but humanoids lag

• Warehouse robots: At least three companies will prominently tout productivity gains from GenAI-driven robotics in their logistics operations

• Humanoid robots: While we might see one or two pilot facilities, scalable humanoid robots remain a more distant prospect
  

Investors in industrial and manufacturing businesses will want to pay close attention to advances in robotics, particularly when planning new facilities.

10. RAG fades as new techniques rise

Retrieval-Augmented Generation (RAG) will decline in prominence as newer "unlimited memory" techniques enable systems to embed and recall documents more seamlessly. A practical prediction: At Bain, we will build at least one tool leveraging document repositories without relying on RAG.

11. No major AI regulation

Despite the hype and fear surrounding AI, 2025 will pass without significant regulatory action in the US —neither an executive order nor congressional act will impose meaningful restrictions.

Conclusion: A Year of Transformation

2025 is shaping up to be another exciting year for GenAI. These predictions are testable, so look for a post next December to see how reliable I am. Investors and companies would do well to monitor these developments closely—those who adapt early to these shifts will stand to gain the most.

What do you think? Which of these predictions do you agree with, and which seem far-fetched? Let me know in the comments, and feel free to share your own 2025 forecasts!

Next week I’ll try to break down the flurry of new releases that came at the end of the year and what they mean for businesses and investors.