The future of AI: 2025 predictions from the AI community and industry leaders

Hey everyone! I’m Ellie, a dev advocate at Pieces, and on January 15th, 2025 I hosted a space on the @getpieces X account with Jim Bennett, Pieces Head of Developer Advocacy, and we had people working in the AI industry (plus a bunch of devs who just love AI) come up on stage and predict all the interesting things that could happen this year in the AI space.

These predictions range from AI and the future of work, the future of GenAI, and the future of AI in business and marketing, to robotics, AI agents, and more!

We asked the speakers first to state whether they feel positively about AI, think it's just a fad, or are more on the pessimistic (anti-AI) side. Then state 1 or 2 predictions regarding the future of AI, particularly the future of AI in 2025.

If you’re interested in listening along while reading everyone’s predictions (and possibly hearing a few predictions that were not included in this post), check out the space on our X profile, and do not forget to check my latest article on LLM trends.

That said, let's get into the predictions 😎

AI’s education revolution & robot cats

Speaker: Ellie (Developer Advocate @ Pieces) 

Stance: Optimistic

Predictions:

AI in education

"I personally am very optimistic about AI. I think some really, really cool things are gonna happen in the future and it’s going to take learning to the next level... My prediction for this year is that AI is going to change the way we learn as a whole. So a lot more apps will have AI tutors—like Duolingo’s AI tutor or CS50’s duck debugger."

Cat Robot (with AI-integration)

“Robot cats. Powered by AI. So they're like your pet, but also your personal assistant. And they can get things for you and remind you what you need at the supermarket. And what if you had a big version at your house that could help you grab stuff? 

But then you can put a small version in your purse, and then it would be like, meow, you forgot to buy ketchup.” (this prediction was made in the latter part of the space) 

The AI backlash & garbage code

Speaker: Jim (Head of Developer Advocacy @ Pieces) 

Stance: Neutral/On the Fence

Predictions:

AI overload

"I think people are going to get really annoyed with the AI and shitification, if you pardon my language there. We’re seeing AI sprinkled on everything. I use a fitness app, and it now has an AI trainer that recommends the next run in the series. It’s this constant AI everything. I think there’s going to be a strong pushback from people."

Failed AI software engineers 

"I'm also not a strong believer in AI as a pilot. So this idea of we're going to replace engineers completely with AI, I think that's destined to fail this year as well. AI will create a whole lot of garbage and then developers are going to have to step in to fix it."

Cost crises & local models

Speaker: Tsavo (CEO @ Pieces) 

Stance: Neutral

Predictions:

The AI сost problem

"The cost associated with these agents is insane. For example, using an IDE like Cursor can run up your bill to $2,000–$3,000 a month. I think I had mine set at $200, and I ran into that bill within a week, which was hilarious. So I stopped using Cursor... I think the cost is something that’s becoming more of a challenge. So I'm pretty excited to see the innovation and driving costs down, particularly around local models. I think a lot of these companies like Cursor, OpenAI, you know, Microsoft, GitHub, they're all going to have problems with cost. And I think that that's going to really open up the space for local small models, which we're big believers in." 

Retrieval over generation

"Our internal roadmap is really focused on not necessarily how do we write net new things or create net new things, but how do we retrieve things that have been created. And I think AI is poised to be deeply helpful in that space. The benefit there is, when you create something from scratch, you have the possibility to introduce problems... versus if you’re retrieving, there’s a lot less room for error on the final outcome. You either find it or you don’t. Nothing bad was generated. Outside of generating stuff, let's look at the other avenues in which AI can be very, very helpful. So, for us (Pieces), long-term memory, we'll see how else other people start applying it."

Agentic workflows & job shifts

Speaker: Sharon (Founder & LLM Whisperer)

Stance: Positive

Predictions:

AI Agents improve

"People are going to get these agentic workflows in check. They’re going to be better"

New jobs emerge

"A lot of companies are saying they’re going to fire people because of AI... But I see new jobs coming to the market."

AI in quality assurance & AI dependence

Speaker: Steven (QA Automation Engineer @ Immunefi)

Stance: Neutral

Predictions:

AI in testing

"AI agents are going to help out in test automation, but I don’t think they’re going to be good at doing it on their own."

Laziness as a barrier

"Laziness will become the biggest obstacle to learning, otherwise known as AI dependence. It’s too easy to just have AI tell you what to write for code... You could fake your way through an interview process."

Human alternatives

"Demand for human or high-touch alternatives to AI are going to increase... We’re going to start seeing an appreciation for human-led stuff rather than AI-led stuff."

Rising restrictions with AI

Speaker: Richard (Technical Writer and React Developer @ Gridiron Survivor)

Stance: Neutral

Prediction:

"I think that AI is going to move towards a direction of companies seeing how far they can go with AI... and we’re going to see the beginning of people starting to try to put some restrictions onto AI, maybe like in the mid-year, near the end of the year."

AI as a tool (not a hammer)

Speaker: Shawn (Senior Developer Relations @ TogaiHQ)
Stance: Pro-AI

Predictions:

 AI agents improve

"I’m hoping that process [of using AI agents] becomes better such that my prompts per minute go up and my results per minute go up... Can it confidently send Steven an email? That’s already something that’s questionable. I wouldn’t go to the next step of code, infrastructure, or the things they pay me six figures to do until it can reliably handle simple tasks like that."

Inclusivity for non-AI users

"How do we make it such that non-AI people don’t necessarily get left behind? My mom is able to use ChatGPT because she happens to have a software engineer son who basically forced her to… How do we include people in this fast moving economy?"

Not a prediction, but I absolutely loved Shawn’s hammer analogy (definitely recommend listening to this part, its around the 27 minute mark):

"If y'all don't know, I am very pro-AI. Like, super pro-AI. Like, the biggest pro of AI. Because I feel like it's very strange for people to be against hammers. Like, I think it's super strange for people to be against, like, screwdrivers. Like, I think it's super strange for people to be against tools. Against what? You’re against hammers or are you against hitting people with hammers? It's like when people have been burning the books. Why are you doing that? The books aren't dangerous. It's the things people do with it. So that's usually what I usually say when people are like, oh, my God, the AI!"

Open Source & Resource Efficiency

Speaker: Krishna (Host @The_404Podcast)
Stance: Pro-AI (with caution)

Predictions:

Lightweight Models

"People will focus more on developing better models that take up less resources... Mistral launched a new AI model, CodeStril. They say it’s two weeks faster than the previous version and proficient in 80 programming languages."

Open Source Growth

"The open-source side will go ahead in the future, taking up fewer resources while performing like closed-source LLMs like GPT models."

Roxy’s prediction (via comment):

“This year is going to be the year where Krishna uses AI to take over the world.”

Commoditization & workspace AI

Speaker: Ali (Senior Developer Advocate @ Pieces)
Stance: Mixed (loves AI as a developer, hates it as a consumer)

Predictions:

AI as a service

"Commoditization of AI is going to happen in 2025, which means let's say if someone is going to GPT today and asking, can I plan a five-day trip to Malaysia or to Europe? It will give you all the information, from portals to let's say flights and so on and so forth. But what if you are able to get like exact recommendations based on the ratings, based on everything and it books it for you? So I feel AI is going to become more of a commodity in forms of agent, in forms of, helpers and so on and so forth… People are going to hire these or use these AI agents or commodities to get their work done."

On-device data

"Letting people manage their own context instead of uploading all of that to the cloud is going to be big."

AI’s impact on marketing

Speaker: Hanna (Head of SEO @Pieces)

Stance: Neutral

Predictions:

NLP-driven content

"I think it’s also going to become more NLP-based on the content we produce. So stuffing the keywords, if they worked before, they’re not going to work now."

Hyper-personalization

"AI tools will anticipate more customer behavior and try to optimize campaigns in real time... All content will be tailored to users’ needs, likings, and preferences."

Marketing automation

"AI will reduce manual effort, enabling one person to do the work of three or four people with ease...if you're a freelancer, I think this is amazing because you can get more clients. You can automate a bunch of your stuff and just really focus more on the strategy or on communicating.. (or) just on having your focus on some high-priority tasks"

Regulation, robots, & a mature ecosystem

Speaker: Francesco (Developer Advocate @ daily.dev)

Stance: Neutral
Predictions:

EU regulations

"There will be more regulations, especially from Europe, about AI."

Mature ecosystem

“I think we will go into 2025 into a more mature ecosystem. This means that maybe the AI apps will start maybe connecting (with) each other"

AI in physical form

"We’ll see AI getting a body—in physical devices and robots. I know this sounds scary, especially if you saw Terminator."

ChatGPT’s decline

"We will see in the whole AI ecosystem a decline of ChatGPT/OpenAI and probably the release of Google Gemini or something else.” 

Single-purpose apps & AI agents

Speaker: Luka (Paid Advertising Lead @ Pieces)

Stance: Pro-AI

Predictions:

Niche AI apps

"Highly specialized, one-purpose apps will explode. LLMs are a bit too general and broad for the average person, which means they don’t really know how to use them. So I think this would be like the App Store boom when the iOS App Store first released. A lot of applications for single purpose users."

Agents become mainstream

"Agents will be incredibly popular. This is also related to the previous point, but the marketing, the positioning there is amazing. So from what I understand, it's not anything especially revolutionary on the technical side, but that name of an “agent” makes a big impact on the average person. It’s like, ‘Wow, it’s an agent—it can do stuff for me.’ That’s just an experience I’ve had talking to people."

AI won’t replace developers

Speaker: Subhra

Stance: Neutral

Prediction:

"AI won’t replace junior engineers... If my code has a bug, it gives the same output again and again."

Open source & wisdom sharing

Speaker: Satoshi
Stance: Pro-AI

Prediction:

"AI will bring a new age of open-source collaboration that’s actually profitable for contributors. I believe that everyone working on open-source with AI—especially engineers—will have the opportunity to replicate their efforts and empower more entrepreneurs through systems that track and record processes, allowing everyone to share wisdom, not just data."

AI for humanity & observability

Speaker: Shivay (Software Engineer & Developer Advocate)

Stance: Pro-AI

Predictions:

AI for all

"AI tools will be adopted by governments to benefit people who don’t have access. I know there will also be a lot of new laws coming into the picture around how this AI data is going to be used. But I certainly feel that a lot of government and a lot of humanitarian efforts will be impacted by AI"

AI observability

"As companies adopt AI agents, they’ll want to monitor costs and performance. AI observability will be huge because every company wants to save costs."

Jim’s conclusion 

“Thank you everyone who's shared their predictions, who's taken time to share their thoughts, feelings, and opinions on this… To kind of sum up, it looks like in general, we're all very kind of bullish about AI as a tool that can help us, whether we think of it as a hammer or whether we think of it as an assistant that's going to support us, someone that's going to build us little apps to support everything we do.

We all like it as a tool, but I think in general, we're all disliking the way humanity is embracing AI, whether it's the (overuse of AI) in apps or whether it's doing bad things or just the talk of replacing people in the workplace. So I think we like the tool. We're worried about people.”

So what’s next?

Hey everyone, it’s Ellie again! Hope you enjoyed reading all of these interesting and sometimes funny predictions. 

What’s next? Well, we’re going to have to wait until the end of 2025 to find out. But even as I was reading over this draft before publishing, I noticed that some people’s predictions were already starting to come true! Especially those regarding new open-source models, less costly models, and new competition popping up (*cough* DeepSeek *cough*). 👀

That said, I’ll see you all here next year to see who was right, and who was wrong. And then I’ll award those imaginary trophies mentioned during the space. 🏆