Here are some (human-generated!) AI developments that we think could happen in AI for business in the coming year…
1) First real-time AI video agents demonstrated
OpenAI’s recent “12 days of OpenAI” Christmas event saw the keenly anticipated arrival of AI video generation service, Sora. The quality of output from tools like this continues to improve and photo realistic creations are very much possible.
Seeing this made us think that perhaps the next obvious step for AI is to produce real-time video avatars to conduct face-to-face conversations on the fly, for example to host video chats. This would certainly take chat bots to another level, although it would be disconcerting not knowing if you are talking to a real person or a bot!
In fairness, Apple actually predicted this nearly 40 years ago in their Knowledege Navigator video (Watch at YouTube)
While we’re definitely closer now than ever its still a big step up from the capabilities of current tools and hence we think we are only going to see this tech as a demo, and probably towards the end of this year.
2) AI in the heart of business applications
Most people think of AI as being accessed through chat style interfaces like Google Gemini, ChatGPT, Claude and Perplexity. However with the availability of straightforward API’s, the ability to use tools, and to output data as JSON, AI will make its way into the core of business application architectures replacing traditional code as a way to achieve incredibly powerful capabilities very quickly.
3) Speech interfaces to IT services become more common
All of the tech is already available on open source for AI-based text-to-speech and speech-to-text input/output so why don’t we see more speech-based business apps? Perhaps because of cultural inhibitions around talking to computers. But equally because the input is more variable than mouse/keyboard controlled data capture and hence presents more challenges to produce the back end logic.
As AI becomes more integrated into business apps the ability to handle variable input will improve and so we expect to see more of these applications, particularly in situations that benefit from hands-free usage – e.g. doctor and vets consulting rooms, salons, engineering benches etc.
And while the older generations may feel shy talking to machines, the coming generations will be much more comfortable with it!
4) First codeless applications start to appear
Provanta recently proved that it’s possible to create a business application with nothing more than a business definition document and an AI client with direct access to database, email, document formatters and other tools. The power and flexibility of the resulting app was extremely compelling, particularly the ability to provide instant answers to any queries, in natural language.
Right now this approach is fine for homebrew applications but will need more purpose built infrastructure before it can be sold commercially, however the possibilities are so tempting that we feel sure we will be seeing apps like this appearing in 2025.
5) Agents arrive but don’t go mainstream
Unless you were living under a rock in 2024 you will have heard about how the coming of AI Agents is going to take the business world by storm. Core to this vision is the concept that agents can take charge of whole processes, making decisions autonomously. We’re not totally disagreeing with this but we think that adoption may take a little longer as people first learn to live with AI embedded in their applications in a non-agentic way. We’re also going to need new tooling that provides strong controls and total visibility over these type of processes
We’ve seen how AI can go off-piste at times and so we feel that while we do start to see these in 2025, it will be a while longer before we give AI agents the green light to start operating autonomously and making decisions on our behalf.
6) AI-led development becomes normal
While we think that codeless applications are the future, its going to be some time before these are viable at commercial scale. Until that comes to pass, the benefits of AI generated code are going to make a huge dent in the traditional software market. As new and updated AI models are being launched with reliable regularity right now, we expect their ability to perform programming tasks to keep on improving.
Its not just the raw coding ability that needs to improve but things like the means to pass context to the AI e.g. to provide the full project background, along with the tools that work with and manage the AI to produce what’s needed.
We expect to see all of these arrive in 2025 and with it a shift from AI-assisted development to AI-led development. This in turn will cause a major reappraisal of the valuation of software – we regularly experience doing a months work in an hour with AI and this productivity gain is bound to have a knock on in the marketplace.
7) AI takes centre stage on the desktop
The currently popular AI services are just the start of our relationship with AI. Our appreciation of what a computer is and how it can help us is going to change markedly as AI becomes more pervasive and starts to work its way into ever more of the devices and services that we use.
Our current expectations of interactions with computers are based on a steady evolution from mainframes to PC’s to browsers and mobiles, from green screen terminals to mice and touch screens. All of which are built around dumb machines that require precise instructions from users and programmers to do whatever it is that we need.
The advent of intelligence will gradually alter our understanding of the relationship between humans and the machines that serve us. As we start to trust them more we will become less task oriented and more managerial in how we interact with them.
This evolution will be facilitated by changes to things like Windows, MacOS, Android which will present an interface that more readily represents the intelligence that it offers. We think we’ll start to see some interesting developments in this direction being shown-off in 2025. (And if you want to see how this could look, refer again to Apple’s Knowledge Book video from 1987 mentioned at the top of this article!)
8) Focus shifts from larger to specialist LLMs
Since the emergence of ChatGPT in 2022, the leading AI providers have vied to create ever bigger and more powerful models to drive their capabilities forward. However, there is a finite limit to this approach which we believe will push them towards improved training methods and better reasoning mechanisms in place of simply devouring ever more data.
Another way that is particularly relevant to AI in business systems is the deployment of smaller more specialised models. This has numerous benefits when AI is working as part of a system, such as improved performance and scalability, and better quality inference with greater consistency.
9) Businesses break free from legacy millstones
So many businesses are stuck on low quality legacy platforms that they have outgrown or no longer match their needs, but which they can’t replace due to the excessive costs and high risks of traditional development approaches.
AI is a fantastic tool for rewriting legacy systems, and this is something we have great experience of at Provanta. It is possible to rapidly recode old systems written in outdated languages, to new systems in current languages. For example, old desktop systems can be converted to web or mobile apps.
Businesses escaping from legacy lock-ins like this will experience a new burst of freedom as well as benefiting from a rise in their business value due to the elimination of the risks presented by legacy platforms.
2025 will be the start of all the old legacy applications being eradicated as AI decisively upends the cost/benefit maths.
10) Talking about AI over dinner is uncool
Its hard to escape from AI at the moment, every email is promoting an AI conference and every other post on Linked-in, YouTube and TikTok seems to rave about it. Its not really surprising given the scale of change coming and the fact we’re all struggling to understand the full impact of that change.
However, maybe 2025 will be the year that we finally stop looking at AI with awe and surprise and it becomes business as usual, much like the internet and smartphones did, all those years ago. Just don’t hold your breath!
Conclusion
The pace of change in AI at the moment is truly breath-taking. Lets see how our predictions stand up in a years time!
