Olya Taran recently joined us at The AI Strategy Summit to give you the playbook for getting team buy-in on AI initiatives. She leads AI adoption at Manulife and has rolled out two AI tools that went from no adoption to over 50% usage in a year.

But this wasn’t happenstance. Olya designed a 3 step framework for driving AI adoption that you’re going to want to copy.

The framework for getting AI adoption

The most important aspect of Olya’s framework: She knew her audience. Rather than forcing the workforce into a one-size-fits-all approach, she met them where they were.

Step 1: Get executives hands-on with AI

Olya started with leadership, because setting the tone from the top is critical. Since executives usually have very little time (and short attention spans), she implemented:

• 15 - 20 minute long sessions
• No presentations – these were hands-on demos tailored to executive use cases

For example, she tasked the executive team with uploading a quarterly report and gave them a challenge: You have 2 minutes to get XYZ information from this report with AI.

The execs had to prompt their own way towards their desired outcome, and experiencing the win for themselves removed skepticism and turned them into instant champions. This buy-in had an authentic and contagious trickle-down effect.

Olya’s pro-tip: “Find the vein”. AKA, insert yourself into existing meetings where leadership will be together rather than trying to play chicken with everyone’s calendars. Otherwise you’ll be waiting forever.

Step 2: Make prompting an essential business skill

Olya realized early on that just telling employees to use AI wouldn’t work because they were missing a key skill: Prompting. To quickly build prompting muscle, Olya orchestrated Promptathons – 45-minute hands-on workshops that are structured like this:

1. Overcome the access issue: Show employees exactly how and where to access the organization’s chosen AI tool (and make it easy to find later).
   
   
2. Let them experience poor outputs: Ask employees to achieve a task and get the effect of a bad prompt first.
   
   
3. Introduce a prompting framework: Now that employees understand the value of good prompts, teach them how to create themin a simple, structured way.
   
   
4. Redo the exercise: Have them attempt the task again leveraging the prompting framework to compare outputs.

These Promptathons became wildly popular with high NPS scores, waitlists, and global demand across the org. They became one of the highest-rated learning sessions at Manulife.

Olya’s pro tip: Let employees see the impact of NOT using a good prompt – then give them the tools to do it well. This will build their confidence when they experience the “aha” moment.

Step 3: Get your “super users” to uncover great use cases

Teaching basic prompting is the (somewhat) easy part – the harder part is getting employees to use AI every single day, for a variety of use cases. To help employees uncover high-value daily use cases, Olya built and leveraged a system of AI super users in each team. From there, she:

• Worked with the super users to uncover and refine novel use cases
  
  
• Coached them on how to create short sales pitches on these use cases – e.g., what business problem it solves, how much time they save, and how they use the tool to accomplish it
  
  
• Ran peer-led sessions where champions demoed their use case, shared the personal workflow it transforms, and challenged peers to replicate it live

These sessions democratized knowledge-sharing and created a sustainable adoption engine by situating AI use in real tasks vs. hypotheticals.

What didn’t work as well

Not everything Olya tried was a hit. Here are two strategies she would approach differently or not bother with at all.

1. Prompt libraries

As Olya put it, prompt libraries were the “fancy new thing” to implement last year. One of her leaders heard about prompt libraries at a conference and requested she try it – but she had misgivings from the jump. Here’s why prompt libraries are mostly skippable:

1. People have to navigate out of their workflows to another tool to use the library. This creates too much work for people to build a habit, and adoption of the library has been very low for this reason.
   
   
2. They’re largely useful in specialized, technical areas. For example, Manulife’s developers using GitHub Copilot love having this prompt library because their use cases are very nuanced and need to be repeatable. Only these pockets of the organization ever use it regularly.
   
   
3. They don’t really inspire new use case discovery. People aren’t browsing this database for inspiration, which means it has very little impact on driving adoption.

2. A decentralized champion network

At first, Olya recruited AI champions from across the org to own the evangelization of AI tools, but it quickly became unsustainable. Her department couldn’t equip thousands of champions with everything they needed to drive peer-to-peer adoption. So instead, she:

1. Shifted to Lead AI Champions: In each office location and each individual department, a lead is chosen to coordinate their own network of local champions. These are typically VP level employees or senior managers.
   
   
2. She runs the program that equips Leads: Olya’s unit provides all materials, presentations, trainings etc. and strategically enables the Leads to use these resources to drive adoption in their area.

This allowed Olya to scale their network of champions to be much larger, but to support those champions in a more strategic way.

Olya’s advice for enterprise leaders

There are a few key learnings that Olya attributes Manulife’s success to:

1. Make AI adoption someone’s job. You can’t leave it to IT or make it a side project for change managers. She says someone needs to be waking up every day and obsessing over the adoption rates and the barriers to entry.
   
   
2. Use right-sized interventions. When getting buy-in, focus on quick wins. 20-45 min workshops are effective for initial AI trainings, and you can scale up to longer workshops as the use cases get more transformational.
   
   
3. Remove all pressure from early stage experiments. Early stage should equal learning stage. Normalize experimentation without expectations and celebrate learning as a valid outcome for pilot projects. Without the fear of failure, you might discover you can do something you never thought possible.