What Changed: AI in Coaching from 2025 to 2026

Prefer to read? Full post below.

Remember when everyone said you needed to master prompt engineering? The promise was simple: learn the magic words, get magic results. But by December, the consensus had flipped. Prompt engineering is dead.

The reason is straightforward. The craft of asking AI the right questions is now being built into applications. You shouldn't need to be an expert at talking to machines—the machines should know how to understand you.

But before that shift happened, we got something else: AI slop.

If you've been on LinkedIn lately, you know exactly what it looks like. And if you're honest with yourself, you've probably made some too. We all got seduced by the promise of effortless content. Some of us even sent out blog posts that were pretty much all AI-generated just to see if we could get away with it. The scary thing? Sometimes people couldn't tell the difference.

The problem is that everyone's AI writes the same way. Generic AI produces generic results, and generic doesn't build trust with clients.

If prompt engineering is dead, what replaces it? Agents. Not chat interfaces where you craft the perfect question, but agentic workflows - AI that handles multi-step processes on your behalf.

Instead of asking AI one question at a time, you give it a task and let it figure out the steps. Want to analyze all your coaching session transcripts from the past month? The old approach meant uploading each one, crafting careful prompts, and copying insights manually. The new approach: tell the agent to find patterns in your sessions this month and let it handle the workflow - pulling transcripts, analyzing them, synthesizing insights. You stop being the prompt engineer and become the task assigner.

The other big shift is from generic chat-based interfaces to special purpose tools. ChatGPT, Claude, and Gemini are like Walmart - big stores with lots of things, but you're lost in there for hours. The future is smaller, specialized stores where you can walk in, get what you want, and walk out. A coaching-specific AI tool knows what session summaries look like. It understands ICF competencies. It's designed for your workflow, not everyone's workflow. Specialized beats generic.

We want AI's power. We're terrified of its cost.

Giving large corporations access to all of our personal information so they can run agentic AI and figure out stuff we find genuinely useful - that sounds scary. And it should. It's not that anyone's worried about somebody reading an email to their mom. The concern is that we want the agentic AI, but we don't want to entrust large corporations with information where they become richer and we just stay the same.

Every time you upload a coaching transcript to ChatGPT, you're making a trade: convenience for data, insights for information about your clients. The 2016 and 2020 elections showed us how aggregated personal information could be weaponized. AI raises the stakes. When you chat with an AI, you're not passively browsing - you're actively revealing your thoughts, your clients' struggles, your professional judgment.

Some coaches are watching the space of local models - AI that runs on your own laptop, keeping all data private. The responses aren't as polished as ChatGPT's, but your data never leaves your machine. Is that trade worth it? For sensitive client information, maybe. You don't need the answer right now. You need to ask the question before uploading.

AI in 2026 is about something coaches have always struggled with: memory across time.

Even coaches with amazing memories sometimes doubt themselves. Did this client say that in session one or session three? What if I'm wrong? It would question my credibility if I called somebody out on something and they said they never said it. AI doesn't forget. It doesn't confuse Client A with Client B. It doesn't wonder if that breakthrough was last month or three months ago. Having that accountability buddy following along with coaching sessions means you can trust yourself to take on 15 or 20 clients. That's not replacing human connection - that's amplifying it.

The uncomfortable part is that most of us don't actually want something holding us accountable. We just want to vent about what upsets us and move forward rather than actually do something. AI accountability works because it reflects back the commitments you've made, and reading that can feel uncomfortable. But once you reframe it, it feels good. Because it's just reminding you of what you said in the past, not telling you who to be. Real growth tools don't just validate you. They hold up a mirror.

2025 was the year we figured out that AI can write - but probably shouldn't write for us. 2026 is the year we figure out how to use AI without losing what makes us human.

That means specialized tools over generic chatbots, agentic workflows over prompt crafting, privacy-conscious choices over convenience defaults, and accountability that serves transformation over comfort. It's not so much the information itself that matters, but your relationship to that information.

AI can surface patterns, remember what you said six months ago, and hold you accountable to your commitments.

But only you can decide what to do with that. The learning goes both ways: you teach AI about your coaching, and AI teaches you about your blind spots.

Want to go deeper?

Michael and I are hosting a live webinar on January 16th: "Stay Human, Embrace AI"—a conversation for coaches who want to use AI ethically without losing their edge.

We'll cover how to evaluate AI tools for coaching, privacy-first approaches that don't sacrifice capability, and the accountability framework that's transforming our own practice.

Bring your questions.

The future of coaching isn't human vs. AI. It's human with AI - done right.