10 Copilot Adoption Pitfalls

I’ve been lucky enough to be involved in several Copilot adoption programs at various stages of gestation and I thought it might be worth reflecting on the most common challenges or mistakes organisations make with their pilot programs:

Lack of Data Protections

We all know that flat file data is a nightmare, folder structures and permissions tend to run wild when not tended carefully and it isn’t unusual to find organisations with data being protected by “obscurity of location” rather than proper permissions, classification, and policy enforcement. Sure, it is possible to exclude repositories from indexing and essentially hide them from Copilot but in doing so you weaken its abilities to assist you. Make classifying and securing your data with Purview or other tools a top priority.

IT or Executive only pilot groups

It’s not, on the face of it, a cheap license and there is a tendency to want to control the scope of the deployment and therefore costs. This is perfectly sensible but if a pilot program is to build a meaningful understanding of how Copilot can impact the business, then the cross-functional cohorts for pilot programs is surely essential?  Many of the lightbulb moments come from encountering how it impacts day-to-day tasks after-all and is the C-suite and the IT team really living every business process?

Lack of Training

It’s not enough just to pop the icon in the right-hand corner and send an email, you need to get people started with some examples and an understanding of some of what is possible. All the research we have also points to a significant improvement in results where prompting skills are enhanced – this doesn’t have to be expensive or in-depth, we don’t need people to understand neural networks, just give them the tools to start interacting more effectively with Copilot.

Failing to share best practices.

This follows on from Lack of Training, bringing people together regularly, particularly in the early stages of a pilot program gives the opportunity to share best practices and learnings. They might spark and idea for a colleague in another part of the business – this is a whole business learning experience – spread the knowledge as fast as possible.

Dismissing Microsoft Graph and Copilot Plugins

The ecosystem for graph plugins that will allow Copilot to access none-Microsoft data repositories is growing rapidly. If your data is in Google Drive or Salesforce or some other systems don’t delay in looking for a plugin to give sight to that data. The more information Copilot has sight of the better it will be able to help your business.

Not Measuring Success

Copilot isn’t a small investment, so any successful pilot program needs to have some measurement of success, not enough attention is paid to structuring a pilot program with metrics or control groups to judge its effectiveness. Even if your business isn’t big enough to warrant a deep statistical analysis even a qualitative assessment process with pilot users has value to help quantify the benefits.  

Treating it purely as an “IT Project”

AI has the potential to re-work business processes, potentially even whole models, as such it requires involvement across the business and needs to be looked at as more than merely an IT project. I’ve seen tension arise between executives and department heads that can see the potential benefits whilst IT leads fear the data security and cost implications to their budgets – I’d argue it’s a special project that needs to be treated as such, otherwise it can easily be stalled.

Moving Licenses between Users too Quickly

I’d have said it takes at least 6-8 weeks for a user to meaningfully adopt Copilot and change old habits – I’ve heard tell that Microsoft were suggesting 11 weeks recently and that doesn’t seem unrealistic either. There is a definite adoption curve with Copilot, and it takes time, sure you spend some time playing with it the moment it springs onto your apps but it is the days and weeks of remembering to “try it” on a task that makes the biggest difference. The licenses aren’t cheap and there is a temptation to buy a very small number and cycle them around users every few weeks, but all this does is generate superficial understanding of its potential.

Focusing solely on Copilot for Microsoft365

There are other Copilots and AI tools on the market, for some users and use cases these will be far better. If you have big user demand and you are worried about shadow IT then your Copilot for Microsoft365 pilot programme can be run in parallel with a deployment of Copilot for the web for example. Would the sales team benefit more from Sales Copilot for example. If its document flow and automation, then Planet AI’s IDA might be a stronger tool.

Not setting Expectations

Microsoft’s demonstration videos show Copilot on its best day. Your results will vary with the quality of the data and the prompting given. And it’s important not to lose sight that this is a young and fast developing product – as I write this Copilot in Excel is still in preview for example, we are not all outsourcing our daily jobs to AI just yet and expect the tools to evolve not be perfect straight out of the box.

In Sum

• Copilot adoption programs need careful consideration, this isn’t a call to “go-slow” but a call to “go-smart” to maximise the value
• Copilot isn’t just another IT tool or project it’s an inflection point in technology of the scale of the internet and web in the late 90’s approach it as such
• If you find yourself needing guidance or support don’t hesitate to reach out to those that have experience in the field – always happy to help!

#CopilotAdoption #AIinBusiness #TechnologyTrends #DigitalTransformation #DataProtection #AIForBusiness #TechInnovation #ITManagement #BusinessProcessOptimization #Microsoft365 #AIAdoption #ChangeManagement #TechLeadership #FutureOfWork #AIIntegration #TechStrategy #BusinessIntelligence #InformationTechnology #EnterpriseTech #TechChange #ArtificialIntelligence