Skip to content

How to implement AI successfully in your organisation: A manager’s perspective

By -

Today's choice: Embrace AI or fall behind:


For businesses today, embracing AI isn't just an option; it's now necessary to stay competitive. Even industries less impacted by these developments (or banning the use) of AI are at risk from employees secretly using services like ChatGPT and exposing company data. Realistically, the only choice for businesses that want to lead, or even to continue succeeding, is to jump in the deep end and embrace AI while equipping your team with the right tools and knowledge to do so safely and effectively. At Oxygen, we've gone deeper than most, figuring out through trial and error how to implement AI in almost everything we do - and the results really speak for themselves.

Below: Oxygen COO Laurent, awarding Claudia, Head of Design, with a certification for prompt engineering.

AI Training featured

 

Our results from AI implementation:

 

Speaking from experience, I have seen the results of AI implementation, and it was everything we had hoped for and more. After running a 6-month trial of GPT-4 post-AI implementation training and certification with our team, we saw the following amazing improvements:

  • 44% reduction in time needed to complete one of our most frequently billed services (from 9 hours to around 4 hours).
  • There was a decrease of 49% in the respondents who said they 'often' feel overwhelmed after AI implementation compared to before.
  • There was an overall net increase in people who felt confident in their colleagues’ ability to create high-quality written content without peer review

The results were clear; efficiency was greatly improved while ALSO improving the quality of work. For more details, please read our full case study here.

How did we get this data, and how do we know it’s accurate? We track all working hours each week within our project management system ClickUp, which allows us to measure, in real time, billable hours attributed to various different tasks. This data exported and compared with results from our employee surveys at the end of this period gave us these results.

 

Sounds good; what’s the catch?

 

Every week, I speak with business leaders, schools and teams about AI implementation; here are some of the most common concerns:

  • What about privacy? Is our data safe?

Yes, you can disable training data sharing if you use either a paid version of ChatGPT or the Enterprise versions via Azure. Meaning your data can be completely private.

  • What about content ownership? Can we get in trouble for using AI content?

Content created by AI is owned by you, and you can use it as you see fit. AI detectors fundamentally do not work, and your concern should be to focus on creating high-quality, non-generic content using these tools. 

  • Won’t the quality of our content/work go down? ChatGPT is super generic

The output will only ever be as good as the prompt and the context you give it. Generic outputs mean your team aren’t using the tools to their full potential.

  • What about when it makes mistakes? Will we be liable for inaccuracies?

As with browsing the internet for references and information now, you, the writer, are responsible for verifying everything written by the AI; this is no different. 

  • Will the content always be on-brand and in our tone of voice?

This depends on how well you can either a) prompt a tool like ChatGPT to understand your brand or b) how well you have set up a pre-prompted instance of GPT-4 in Microsoft Azure or other tools.

  • How can my team access these tools? Don't we need a VPN?

Depending on the tool you decide to use, you might not need a VPN. GPT-4 via Azure and Microsoft Teams can be available to teams globally without needing a VPN or overseas phone number/credit card for account registration.

 

The steps of AI implementation:

 

  1. Choose the right tool: Ensure you understand which tool is best for your business. If you are in a region where ChatGPT isn’t available, or you are a larger organisation with multiple departments, it’s probably time to consider GPT-4 via Azure, Microsoft Teams or other options like HubSpot.
  2. Team Training: This is one of the biggest failure points for new businesses using AI that we see. They aren’t experts on using the tool, and their team either has a bad experience or becomes lazy using AI to create generic results.
  3. Team Certification: The gamification of our AI training was a huge success, and the team felt very accomplished and proud that they could achieve something to showcase their hard work in learning a new tool.
  4. AI Compliance Documentation: This is essential for your team to sign off on, confirming that they understand the risks of using AI, and at the end of the day, they alone are responsible for the output. 
  5. Public AI Policy: Your clients and customers need to know how you use AI and what precautions you take to keep their data safe.
  6. Custom Prompt Engineering: GPT-4 does better with more succinct prompts, so condensing your brand information and context into clear prompts is essential for getting good results on the first generation. 
  7. Custom Chatbot Development: Taking things to the next step, this is creating a chatbot either per department or purpose that is already pre-prompted and requires less prompt knowledge.
  8. Surveying Your Team: If done correctly, this is a huge change in how your team works. Surveying your team before, during and after implementation will help you understand how this impacts your team and your business.

 

Buy-in from management

 

The key to success in this initiative is change management and the commitment from management towards proper AI adoption. The tone of communication is also crucial, as employees often worry about AI replacing their jobs. It should be clearly conveyed that AI is meant to enhance efficiency, freeing up the team for tasks that AI cannot perform.

Organizations frequently seek the efficiency benefits of AI without fully committing to understanding the technology. This knowledge is essential for leading teams and managing any less-than-perfect outcomes. If a team inexperienced with AI isn't using it properly, management must be able to detect when the output isn't up to par and determine the root causes of these issues.

Need support in implementing AI for your business? Speak with one of our experts for a free 15-minute consultation on how your business can use AI to become a digital leader.

Laurent Ross

About the Author

15+ years of work in client and project management in China, Hong Kong, Singapore and Thailand - Laurent is uniquely qualified to manage campaigns which require a deep knowledge of both Western and Chinese digital systems.

Find me on:

You may also like

Mastering Prompting on ChatGPT
Read More
The Best AI Image Generation Tools For Your Business in 2024
Read More
The 3 Best AI Voice Tools For Blog Posts
Read More