Everyone can use AI, not everyone is good at it

Everyone can use AI

Anyone can open Claude and get an answer in seconds. Anyone can turn on an AI feature in Salesforce. That's the promise, and it's true.

But being able to use a tool and being good at it are different things. Give the same oak planks to a master floor installer and to a first-timer with a rented saw. One gets a parquet floor that lasts a century. The other gets gaps, splinters, and a room nobody wants to walk in.

AI works the same way. The people who are most proficient with it use it to build remarkable things. The people who aren't often create more mess and confusion than they had before they started.

What AI slop looks like

AI slop is what happens when AI gets thrown at the wall to see what sticks. It's usually well-intentioned, and nobody likes the result. The pattern is familiar:

• Licenses without a plan: seats get bought, a few enthusiasts try things, and nobody owns the outcome.
• Answers from messy data: AI connected to a CRM full of duplicates and stale records, producing confident answers built on bad information.
• Access nobody reviewed: AI inheriting permissions that were too broad to begin with, and surfacing data people were never meant to see.
• Edit access in the wrong hands: AI making changes in a Salesforce org that nobody with real admin experience is checking.
• Every tool, a different answer: one team asks Claude, another asks ChatGPT, a third uses Gemini or Perplexity, and each gets a different "revenue" or "active customer" because nobody wrote down what those fields mean or which ones are actually maintained.
• Output nobody measures: lots of activity, lots of generated content, and no one able to say whether anything got faster, cheaper, or better.

The result is more work, not less: more content to review, more records to clean up, and more people who decide "AI doesn't work here."

What polished AI looks like

Polished AI starts from the business, not the tool. Before anything gets built, someone takes the time to understand how the work actually happens today, where it breaks, and what would be worth fixing.

• Discovery first: map the process, the systems, the data, and the people involved, and measure how long the work takes now.
• Scope one problem: pick a specific job with clear rules, a clear owner, and a number to move.
• Clean the foundation: fix permissions and data hygiene so the AI sees the right things and the things it sees are true.
• Write the data dictionary and query guide: document what every important object and field means, which fields are trustworthy and maintained, and how to ask for the numbers that matter. Then make that guide the standard for every AI tool in the company, so Claude, ChatGPT, Gemini, and Perplexity all answer from the same, true definitions.
• Build an agent for that job: not a general-purpose assistant pointed at everything, but an agent designed, tested, and limited to the problem it solves.
• Enable the people: onboarding, guides, and support so the team actually uses it.
• Measure the after: hours saved, turnaround time, cost, and the work people now have time for.

Polished AI is quieter than AI slop. There's less noise and fewer demos. There's just work that used to take hours and now takes minutes, and people spending their time on the parts that need judgment.

Proficiency is the multiplier

AI doesn't replace expertise. It multiplies it. An experienced Salesforce admin working with Claude can do the work of several admins, because they know what good looks like, what a change will break downstream, and when the AI's answer is wrong.

In skilled hands, AI multiplies good work. In unskilled hands, it multiplies mess.

The same multiplier works in reverse. Someone without that experience gets the same speed, and uses it to make mistakes faster and at greater scale. That's why the most important AI decision a company makes isn't which model to buy. It's who sets it up, and how.

Vision without execution is hallucination

Every company has an AI vision now. Very few have AI doing measurable work. The gap between the two is execution: discovery, scoping, clean data, careful permissions, enablement, and measurement.

That's the work we do. We're operations specialists, and we treat an AI implementation the way a craftsman treats a parquet floor: measure twice, fit every piece, and polish it until it holds up for years.

If you want to start with the foundation, our AI Readiness Audit covers permissions, data hygiene, and a plan for where AI will actually pay off.