This post was also published in Inc.
Everyone talks about AI, but getting your own organization to
meaningfully use it can be frustratingly difficult.
Until now, specialized personnel have handled tasks like SEO keyword identification, content writing, and image editing. If you expect to consolidate those responsibilities to a single person or project lead using
AI tools, you’re dramatically increasing mental load. Instead of collaborating with and delegating to a team of specialists, they now have to conceptualize the deliverable, precisely prompt the AI, and refine the output.
But why can’t agents or automations just replace the specialists?
Well they can. But understand that automating multi-step organizational tasks, such as generating outbound sales messages, requires a structured approach, not just a simple directive to “start using AI.”
Here are four practical and proven approaches to getting through implementation blocks:
Get an AI Guy
Automating business processes is a specialized skill well beyond most employees, so bring in a software engineer or a business process analyst who can collaborate with the team to thoroughly understand and automate the repetitive steps involved. Meanwhile a consultant for targeted staff training accelerates learning and gets people comfortable with new workflows and tools. There’s nothing like a sherpa to show people how to climb a mountain—people like learning from people.
Document Effective Prompts
If we went back in time two years and told people to tell an LLM “you are a data scientist at a prestigious university,” they would have thought it was lunacy. The good news is that after a couple of years there are numerous examples online of how to prompt an LLM and sometimes the LLM itself can suggest great prompts.
So if you need to get something done, look up what to prompt rather than just typing in what comes to mind. For example, here are some
Google Nano Banana image prompts that produce professional quality images. You don’t know what an F-stop is, but the Nano Banana will know what to create.
Give the AI More Context
The more an AI knows about your business, the better it can perform. Rather than ChatGPT, consider using the new AI features offered by your existing software vendors.
For example, Salesforce has launched a slew of new AI features that makes tasks like writing a followup email for an existing customer much easier because Salesforce already has important contextual data, like notes from the last meeting and upcoming renewal dates. Both Microsoft and Google have also added extensive AI file capabilities so you can load a lot of your existing documents into the Microsoft Office 365 and Google Workspace AIs, and staff can start asking questions based on the materials.
Set Cost Controls and Guardrails
Most AI tools charge by nebulous “tokens” and it is notoriously difficult for employees to understand how much AI tasks cost. For each AI tool, set a spending limit or a trigger to notify if a spending threshold is exceeded. And indeed, sometimes you do want Hal to not open the pod bay doors. Most AI tools have “guardrails” where you can specify how to handle sensitive data and what is and is not allowed.
With that, unleash your staff with AI this year. They’ll have fun and your organization will be more productive. Autobots, roll out!