Stop Teaching AI Prompts. Start Building AI Workflows.

We run AI training sessions for a pharmaceutical teams.

The biggest takeaway isn't about prompting techniques or which buttons to click. It was about everything that happens before the training even starts.

If you're a company thinking about upskilling your sales team on AI, here's what I've learned — from running A.I programs.

Start with the problem, not the tool

Before we walk into the training room, we've already sent out a questionnaire, worked with the I.T Team, worked with real examples, and understand the companies products and individual and team goals.

For a sales team We wanted to understand what actually eats into a sales rep's day. The answers were predictable, but hearing them quantifies was valuable:

• Travelling between appointments
• Writing and responding to emails
• Producing call reports and territory summaries
• Scheduling and rescheduling appointments
• Planning meetings and preparing agendas
• Reading clinical information, product updates, and market data

We also gauge their comfort level with AI. Some have dabbled with ChatGPT at home. Others hadn't touched it. A few were quietly sceptical. All of that shaped what we built for them.

The point is: we didn't show up with a generic "Intro to AI" slide deck. We showed up with their problems already mapped to their tools.

Microsoft Copilot is seriously underutilised

The last team we trained had enterprise Microsoft 365 with Copilot access already included in their licensing. Most of them didn't know that. A few had seen the icon but never clicked it.

This is more common than you'd think. Companies are paying for Copilot and barely scratching the surface. It sits inside Outlook, Teams, Word, Excel, and PowerPoint — the tools sales teams already live in. You don't need to convince anyone to adopt a new platform. You just need to show them what the one they already have can actually do.

We walked through practical scenarios: drafting follow-up emails from meeting notes, summarising long email threads, pulling together weekly reports from scattered data, and preparing briefing documents before a call. Their reaction wasn't "that's cool" — it was "why hasn't anyone shown us this before?"

The real lesson: training isn't enough

Here's the uncomfortable truth. You can run the best AI training session in the world. You can teach people how to write great prompts. You can explain what large language models are and aren't. You can get the room excited.

And six weeks later, most of them won't be using it.

Not because they don't see the value. Because they're busy. Because writing a prompt from scratch every time they need a call report feels like extra work. Because they're not sure if they're "doing it right." Because the gap between understanding AI and embedding it into a daily workflow is larger than a two-hour training session can bridge.

What actually drives adoption

After running many of these sessions, I've landed on three things that make the difference between a training that sticks and one that fades.

1. Build a shared prompt library before you train

Don't ask twenty sales reps to individually figure out how to prompt AI for a call report. Write the prompt once. Test it. Refine it. Make it available to everyone.

We now spend significant time before the training building out prompt templates tailored to the team's actual workflows. Email templates. Report structures. Meeting prep frameworks. By the time we walk into the room, we're not saying "here's how to write a prompt." We're saying "here's a prompt that already works — try it with your data."

The shift from "learn this skill" to "use this tool" is everything.

2. Get IT in the room early

This one catches people off guard. We've had sessions where a rep tries to use Copilot in Teams and gets a permissions error. Or where company policy blocks certain data from being processed. Or where the security settings haven't been configured to allow Copilot to access SharePoint files.

Now we meet with IT before every training. We walk through permissions, data governance, and security policies. We make sure everything actually works before a single rep sits down. Nothing kills enthusiasm faster than "sorry, that feature isn't enabled for your account."

3. Don't leave adoption to the individual

This is the big one. The companies that get real value from AI aren't the ones with the best-trained employees. They're the ones that build AI into the infrastructure.

That means pre-built templates in shared drives. Automated workflows that trigger without anyone thinking about it. Copilot prompts embedded into existing processes rather than bolted on as an extra step.

Your sales team's job is to sell. Every minute they spend learning to be an AI power user is a minute they're not spending with a customer. So do the hard work for them. Build the workflows. Set up the automations. Hand them a system that makes the high-value activity — being in front of customers — happen more often.

The bottom line

AI training is necessary. But it's not sufficient.

If you want your sales team to actually adopt AI, you need to move beyond education and into implementation. Pre-build the prompts. Pre-configure the tools. Pre-solve the IT issues. Make AI the path of least resistance, not an additional task on an already full plate.

The companies that understand this will have sales teams that are measurably more efficient within weeks, not months. The ones that don't will have a room full of people who attended a great training session and then went back to doing things the way they've always done them.

Don't just teach your team about AI. Build AI into the way your team works.

---

Matt Martin is the founder of Medware Group, an AI-native healthcare technology company based in Sydney, Australia. He works with pharmaceutical and healthcare organisations on AI adoption, automation, and digital transformation.