Ever been in a situation where you returned to work after a 4-week leave, only to realize everything has changed?

It could be new faces, furniture shifts, or that your favourite snacks are no longer stocked in the pantry!

It takes a while to adjust to these changes.

The rate of change for AI is like that… times 100!

If you blinked, you might’ve missed it.

In just the last 90 days, the AI landscape didn’t just shift—it erupted. We’re no longer debating if AI will change our industries. It already has. What we’re seeing now is the next wave—one that’s faster, cheaper, and a whole lot more autonomous.

From China’s surprise power move with DeepSeek, to the quiet but seismic rollout of “Agentic AI” that can act on your behalf, Q1 2025 has felt less like a quarter and more like a sci-fi fast-forward button got pressed.

This isn’t just another trends report. It’s a field guide for corporate leaders and decision-makers who want to stay relevant in a world where AI doesn’t just assist—it acts.

From a list of 20+ highlights, I’m going to stick to our tradition of five. (Yes, in case this is new for you, at Radiant Institute, every 3 months, we take stock of the 5 biggest trends in the field of AI. If you want to check out our previous issues, here are the links for the last 2 issues, Q4 2024, and Q3 2024.

Ok, enough talk. Let’s break down the five biggest shifts from AI’s wild Q1.

Now everyone can have their individual research assistant!

AI Highlight #1. Deep Research tools are no longer assistants — they’re analysts.

Let’s talk about Sarah, a marketing director in an insurance company. She is asked to present to the board of directors on the state of the insurance industry. In the past, she would either spend hours in the evening in front of a computer, fueled by a cup of coffee, and deep dive into dozens of industry reports, Google tons of articles, open multiple tabs… just to get enough data for her presentation.

That’s probably a world before 2025. Today, she can just enter the prompt into an AI tool, refill her water bottle at the pantry, and a 10,000+ word report, complete with all the references and citations waiting for her to review.

With Deep Research functionality, Perplexity, OpenAI, and Gemini didn’t just upgrade in Q1 — they graduated. These tools are now your smartest colleague, capable of producing high-quality research in minutes with citations, structure, and contextual awareness that would rival your best intern or even junior consultants.

• Perplexity’s Deep Research hit 93.9% factual accuracy on benchmark tests. That’s not just fast — it’s trusted. Better yet, it processes reports in just 3 minutes, compared to ChatGPT Pro’s 27. Speed is the new credibility.
• OpenAI’s Deep Research, powered by the o3 model, solved unsolved ARC puzzles, performed deep UI/UX audits, and handled reasoning tasks with multi-source citations. It’s less “assistant” and more “autonomous analyst.”
• Google’s Gemini Flash 2.0 introduced deep reasoning chains and query memory, allowing users to have actual iterative conversations around research goals. That’s a game-changer for strategy and trend mapping.

These tools don’t just summarize—they analyze, compare, cite, and recommend. They’re becoming your go-to for strategy decks, trend reports, and due diligence.

And the pricing structure? Perplexity remains competitive due to its open-source foundation. OpenAI charges $200/month for power users, while Gemini has opened up limited usage for free — a move likely intended to drive mass-market traction.

💡 Bottom line: This new breed of AI isn’t here to assist. It’s here to collaborate, iterate, and outpace. Now, everyone can have an assistant at work (and frankly, why stop at one?)

China’s AI is making waves the size of tsunamis!

AI Highlight #2. DeepSeek & China’s AI Play: From Challenger to Champion

If 2024 was about catching up, Q1 2025 showed China taking the lead. Quietly. Boldly. Unapologetically.

• DeepSeek R1, trained for just $6M, achieved performance within striking distance of GPT-4—a model that cost OpenAI over $100M to train. It wasn’t just cheaper. It was more strategic.
• The model became the most downloaded free AI app in the U.S. App Store in January 2025—a clear signal that global users were ready for an alternative.
• Its open-source sibling, DeepSeek R1 1776, dropped censorship constraints and spread like wildfire among developers and startups. Within weeks, Tencent integrated it into WeChat, supercharging its search and smart assistant capabilities for over a billion users.
• And the geopolitical fallout? $593B in U.S. tech market cap erased in under a month.
• Nvidia’s stock took a 17% dive. Why? Because faster, lighter inference tools are reducing demand for GPU-heavy infrastructure.

But here’s what matters: DeepSeek is energy efficient, running on 1/10th the energy of comparable Western systems. And it’s backed by homegrown chips from Huawei’s Ascend line, skirting U.S. semiconductor sanctions.

This isn’t just a tech story. It’s a geo-economic pivot, and it sends a clear message: China isn’t waiting to be invited to the AI table. It’s bringing its own.

💡 Bottom line: DeepSeek didn’t just close the gap. It redefined the race. Venture capitalist and co-founder of a16z, Marc Andreessen, sees the release of R1 by a Chinese AI lab as the “Sputnik Moment”, and drags comparison to the Soviet Union’s launch of the first satellite into orbit.

The next phase of “software”.

AI Highlight #3. Agentic AI: When AI Stops Suggesting and Starts Doing

A few weeks back we talked about the software evolution.

In a nutshell, it all started with Applications, with tools like Microsoft Office, Adobe Suites and Skype (which will be missed next month), etc.

And then, in November 2022 when ChatGPT made it’s debut, we entered the era of Assistants, “software” that is smart enough to write emails, sketch images, and bring you up to speed even if you join a Zoom meeting 15 minutes late.

Now, we’re in the era of Agents – autonomous, or semi-autonomous AI tools that can think on it’s own, and get the work done with minimal human supervision. Very similar to the Deep Research feature we talked about in AI Highlight #1.

Agentic AI is where things get wild, because we’re talking about AI that doesn’t just respond to your queries—it takes action on its own.

• OpenAI’s Operator lives in a secure browser environment and can autonomously book travel, buy products, and interact with web interfaces. It’s your AI executive assistant — with a credit card.

• Manus, China’s rival, created and completed a full academic research study in under an hour: survey design, data gathering, analysis, visualizations, and a final write-up.

• These tools rely on hierarchical agents — planners, doers, and reviewers — collaborating to execute multi-step goals.

• The financial world noticed. Agentic AI mentions rose 275% in Q1 earnings calls, overtaking chatbots and copilots.

• OpenAI projects Agentic AI could handle 5–10% of economically valuable tasks by the end of 2026.

The promise? 10x productivity.

The threat? 10x complexity and oversight risks.

If you look up over the horizon, you can see the privacy hawks are already circling. Operator’s browser-level control has raised red flags about data exposure and decision transparency. Meanwhile, education leaders are scrambling to address tools like Manus, which could write entire university assignments undetected.

💡 Bottom line: We’ve built agents. Now we need ethics, governance, and guardrails. As Jensen Huang rightfully said, we need IT managers to function like HR managers. The only difference is they are not handing people, but an army of agents.

Swimming in funds is fun!

AI Highlight #4. Record AI Funding: The Boom Just Got Louder

Say all you want about the economic state of world, we cannot deny that money isn’t just flowing into AI. It’s gushing, like a broken reservoir.

• In Q1 alone, AI startups raised $10.4B, eclipsing 2024’s Q1 by a full $2B. xAI took $10B. Figure AI raised $1.5B. Anthropic closed $3.5B at a staggering $61.5B valuation.

• AI now accounts for 20% of all global VC deals, tripling fintech’s share.

• The money is moving from LLMs to vertical-specific AI: radiology tools, synthetic voice models, legal automation, autonomous robotics.

• Real-time translation and synthetic media pushed Voice AI funding to $500M this quarter alone.

• Cloud players like Microsoft and AWS poured $40B+ into AI infrastructure, doubling down on partnerships with OpenAI and Anthropic.

Investors have moved past the shiny object phase. They’re now looking at compound use cases — where AI becomes embedded into the core fabric of industries.

And the impact of the funding is obvious. More marketing campaigns that target the masses (yes, there are billboards, signages and even Superbowl ads on AI companies now), faster model releases, and most importantly, free usage even for non-paying users!

Just last week, an AI feature that took over your social media feeds are the anime-like photos of your friends and colleagues, in the style of Studio Ghibli. If there’s one company to “blame” for turning your socials into an anime fan fest, it’s OpenAI’s new image generation feature.

Features like these used to be kept behind paid walls, and the investment paid off. OpenAI added one million new users in a single hour after introducing the feature, a number that they took 5 days to achieve back in November 2022!

💡Bottom line: We’re not in a bubble. We’re in a build phase. If you think the AI revolution wave leading up to now is crazy, 2025 will definitely overwhelm you. Because in my years working with startups, when funding comes in, two things grow – brand awareness, and teams. Which leads us nicely into highlight number 5.

The next frontier will be fought around AI-enabled talents!

AI Highlight #5. The Talent War: More than Salaries, It’s a Survival Game

If AI is the rocket, talent is the fuel. And right now, everyone’s fighting over the last few canisters.

• Senior engineers are raking in $450K–$650K, while companies are dangling 70% equity offers for top researchers.

• Nvidia, AMD, and Meta boosted pay packages by up to 28% to keep their top architects from being poached.

• Remote-first roles declined 40%, as firms like OpenAI now demand in-office collaboration to protect IP and train agentic models.

• China’s AI hubs, like Hangzhou and Shenzhen, are luring U.S. expats with tax incentives, research grants, and 10-day visa approval programs.

• Meta, Google, and Amazon launched AI mentorship and internal upskilling programs to reduce attrition—and it’s working. Attrition dropped 22% in pilot teams.

Meanwhile, startups are getting scrappy: offering liquidity access, asynchronous work structures, and even co-authorship on research papers.

There are even trends where candidates would prioritize a company’s rules on using AI for work, rather than the ability to work from home! And if a company restrict the usage of AI, they know there are probably a few competitors who are more than happy to take them in (with paid, premium accounts nevertheless!)

The shift is slowly reflected on job descriptions as well, where employers are adding a line “Must be able to use AI tools”, right after the classic work productivity tools like Microsoft Excel and the gang.

💡 Bottom line: The new job title isn’t “AI Engineer.” It’s AI Differentiator. AI is reshaping how we recruit and retain talents, and on the flip side, talents who are not AI-enabled will be treated as resumes without photos!

What will the next 90 days look like?

Final Thought: Don’t Just Watch the Shift — Be the Shift

Let’s step back for a second.

In just one quarter, Deep Research tools became the smartest colleague you never hired.

China flipped the global narrative with DeepSeek, turning disruption into dominance.

Agentic AI stopped asking and started acting. Investors rewrote their playbooks — not for hype, but for long-term bets.

And the talent market? It’s no longer about finding the best. It’s about being the most AI-enabled.

This isn’t a trend you can afford to bookmark for later. The “wait-and-see” attitude will not serve anyone in the rapid AI world we live in.

The people who win in this new cycle aren’t just staying informed. They’re retraining their teams, rethinking workflows, and reimagining what “efficient” really looks like.

Because in 2025, you don’t get points for noticing the shift.

You get rewarded for becoming it.

Until the next quarterly update, here’s us wishing you success in your AI enablement initiatives!