With AI, show is always, always, always better than tell.
It's a new concept. It's a vague term. And it means different things to different people. So it's no wonder clients and management teams are sceptical, worried, or hesitant.
They need to see, with their own eyes, how AI can:
- Do the task they're interested in — to a reasonable quality, similar to that of someone with a few years of relevant experience
- Do the task consistently, safely and not create new problems — like making up facts or being wrong
- Do the task quickly and cost-effectively — it can't cost the earth
- Drive genuine productivity improvements in their teams — by removing basic tasks from people's jobs and letting them work on harder, more impactful problems
With those points in mind, you need to:
1. Choose the task you think AI can have a large impact on, e.g. drafting social media posts or responding to customer support requests
2. Create a simple PoC, using a No-Code app builder or @foreverhumble74. Think about the inputs required, e.g. a customer support email. Think about what's needed by the AI to complete the task, e.g. your help docs, or customer support FAQs
3. Simulate 5-10 examples of this task, where the output is created by AI, e.g. using ChatGPT or your AI PoC to draft email responses to 5 real customer support queries
4. Show a comparison of the results: output, quality score, time taken, $ saved
5. Show your findings, but set expectations that it won't be perfect or ready to launch. If you can show measurable results from a PoC that took a few hours or days, imagine how good the results could be if you or your team dedicated more time to it.
Final point, spend 30 mins reading about how AI models (specifically Large Language Models) work. What they're good at and what they're bad it. You'll be in the top 1% of your company in terms of AI knowledge. This will also help you field questions like "what it writes a rude or dangerous support email reply?".
Don't give up. Using AI within every company is the future, whether your boss or client likes it or not.
It's a new concept. It's a vague term. And it means different things to different people. So it's no wonder clients and management teams are sceptical, worried, or hesitant.
They need to see, with their own eyes, how AI can:
- Do the task they're interested in — to a reasonable quality, similar to that of someone with a few years of relevant experience
- Do the task consistently, safely and not create new problems — like making up facts or being wrong
- Do the task quickly and cost-effectively — it can't cost the earth
- Drive genuine productivity improvements in their teams — by removing basic tasks from people's jobs and letting them work on harder, more impactful problems
With those points in mind, you need to:
1. Choose the task you think AI can have a large impact on, e.g. drafting social media posts or responding to customer support requests
2. Create a simple PoC, using a No-Code app builder or @foreverhumble74. Think about the inputs required, e.g. a customer support email. Think about what's needed by the AI to complete the task, e.g. your help docs, or customer support FAQs
3. Simulate 5-10 examples of this task, where the output is created by AI, e.g. using ChatGPT or your AI PoC to draft email responses to 5 real customer support queries
4. Show a comparison of the results: output, quality score, time taken, $ saved
5. Show your findings, but set expectations that it won't be perfect or ready to launch. If you can show measurable results from a PoC that took a few hours or days, imagine how good the results could be if you or your team dedicated more time to it.
Final point, spend 30 mins reading about how AI models (specifically Large Language Models) work. What they're good at and what they're bad it. You'll be in the top 1% of your company in terms of AI knowledge. This will also help you field questions like "what it writes a rude or dangerous support email reply?".
Don't give up. Using AI within every company is the future, whether your boss or client likes it or not.
