Stop Using One AI Tool for Everything

Most cleaning and restoration business owners who have dabbled in artificial intelligence have done it the same way: pick one tool, use it for everything, and wonder why the results feel off.

Dean Mercado, founder of Online Marketing Muscle and a longtime AI strategist for the cleaning and facility services industries, said that’s the wrong approach.

“Everything in your business should start with a challenge that you have, a problem you need to solve, something you want a little help with, and then you choose the right tool for the right job,” Mercado said. “I’m not going to use a screwdriver to pound a nail in. I mean, I will if I have to, but I shouldn’t if I don’t have to.”

Mercado outlined seven core business challenges, along with the best AI tools for each, based on his research and personal daily use.

Before diving in, he offered one caution: always vet the companies behind the tools you use and be careful about what proprietary or sensitive information you share.

Writing: Claude leads the pack

For drafting proposals, SOPs, employee reviews, client responses, or any content where the human voice matters, Mercado ranked Claude as the best option, with Gemini as a solid middle choice and ChatGPT as a safe, if generic, fallback.

“Claude seems to get the human side of the voice better than the rest,” he said. His technique: draft in one tool, then run that draft through a second to grade and improve the result. “I’ll pit one against the other—here’s what I got from ChatGPT, what do you say about this? How can I make this a 10?”

Research: control your source material

For general web research, Mercado recommended Perplexity over a standard Google search. But when the research needs to draw on specific, trusted sources—industry standards, compliance documents, proprietary training material—his top pick was Notebook LM, a free Google tool that lets users build a controlled knowledge vault by uploading their own documents and URLs.

He also flagged Grok, which draws on data from X (formerly Twitter), as a useful wild card for picking up on industry sentiment and conversations that other tools might miss.

Planning: let Gemini connect the dots

For daily, weekly, quarterly, or annual planning—including route planning and crew scheduling—Mercado placed Gemini at the top, largely because of its native integration with Google Workspace. ChatGPT ranked as a solid good option, with Claude closing the gap as its planning capabilities have recently improved.

“Because of the way it already taps into everything I use in Google Workspace—my email, my calendar—Gemini is a little bit more powerful,” he said.

Strategy and brainstorming: ChatGPT’s memory stands out

When it comes to thinking something through—pressure-testing a decision, pricing a new commercial account, mapping out options—Mercado called ChatGPT his go-to, citing its stronger memory across sessions as the differentiator.

“If I want help thinking, not just help typing, I go to ChatGPT,” he said. He will often cross-check by running the result through Claude for a sanity check.

Numbers and data: Gemini again

For job costing, margin analysis by service type, recurring versus one-time revenue breakdowns, or any task that involves turning raw numbers into usable insight, Mercado pointed to Gemini—especially for users who store data in Google Sheets, where Gemini can connect directly.

“If you don’t know your numbers, you don’t know your business,” he said. “And it’s not just enough to parrot the numbers back to me. Knowing what they mean is a different story.”

Visuals: Nano Banana, then Canva

For social graphics, website imagery, before-and-after photos, recruiting visuals, or ad concepts, Mercado said Nano Banana—Google’s AI image generation model built into Gemini—delivers the strongest results when accuracy to a prompt matters. Canva AI ranked as the better option for refining and finishing images, and ChatGPT’s image function came in as the good baseline.

His workflow often combines both: generate the image in Nano Banana, bring it into Canva to remove or replace elements, and add text or branding. “The image is so realistic, you would not know that it was AI generated,” he said. “You’ve got to know how to prompt—if you don’t know how to ask for exactly what you want, you’re going to get whatever it thinks you’re saying.”

Learning: build your own training room

Mercado’s top pick for accelerated learning was again Notebook LM, which he uses to create targeted “training rooms” by loading in source materials—videos, newsletters, SOPs, industry standards—and then querying the notebook to extract exactly what he needs. YouTube ranked as the better option for general learning, and Google Search as the baseline good choice.

He described using this approach to help clients avoid expensive courses by building a Notebook LM with the best available material on a topic and interacting with it directly. “You could do that in 15 minutes, and you’ll have what you need,” he said.

The same method works for staff training: load in company training videos, SOPs, and process documents, and the resulting notebook draws exclusively from that internal content rather than from outside sources.

Three honorable mentions

Beyond the seven core challenges, Mercado highlighted three additional tools for owners who think faster than they type, run a lot of meetings, or want to automate repetitive tasks.

For voice-to-text dictation, he recommended Wispr Flow (wisprflow.ai), a cross-platform app that transcribes speech into polished text across any application. “I want you to imagine you’re in the field and you want to communicate something,” he said. “Pop open Wispr Flow, dictate whatever it is you want to talk about, and then keep moving.”

For meeting notes, he listed three options at different maturity levels: Otter.ai for a tried-and-true solution, Fathom for its free version and easy sharing features, and Granola as the emerging choice for users who don’t want a bot joining their calls. Unlike the others, Granola captures audio directly from the device without appearing as a participant.

For automation, Mercado’s default combination remains Keap (formerly Infusionsoft) paired with Zapier—a stack he said can automate nearly anything short of physical cleaning work. For owners ready to go deeper, he pointed to Make and n8n as proven platforms for building more complex automated workflows.

Start with the problem, not the tool

Mercado organized AI tools into three broad levels: chat (conversational tools like ChatGPT, Claude, and Gemini), automation (tools that execute tasks on command), and agents (AI systems that independently pursue an outcome using whatever resources they need). Most cleaning and restoration owners, he said, are still in the chat tier—and that’s a legitimate place to get significant value.

His parting advice was to resist the temptation to get married to any single tool.

“You’re not married and you shouldn’t be married to one tool,” he said. “If you’re not getting what you want from one, switch. Or use the technique I do: pit one against another. It gets easy. Trust me.”

Watch the interview and listen to the podcast: