Kick Point founder Dana DiTomaso’s answer to how she uses AI comes down to one idea, repeated across her whole conversation with BrightLocal co-founder Myles Anderson.
AI isn’t there to finish the work. AI is there to help you get unstuck or serve as a kick point (sorry) at the beginning of a project.
Dana’s spent 26 years in digital marketing, starting in SEO and local SEO before analytics became her focus. She’s the founder of the agency Kick Point and runs Analytics Playbook, a training platform for agencies working in GA4, Tag Manager, and Looker Studio. As well as the excellent host of our Google Analytics 4 course on BrightLocal Academy.
She sat down with Myles to talk through where AI actually earns its place in her work.
Traffic isn’t the whole story anymore
“In the past, success equaled traffic. And now, of course, with AIs and LLMs and everything else, traffic may not be the thing that determines success anymore. And I think that this means that SEOs are having a really difficult time.”
Success equals traffic was the model SEOs built their reporting around for years, and AI Overviews and LLM answers are breaking it, without a clear replacement yet.
Dana raises two candidates to replace it, but acknowledges they’re both still unsettled.
- Citations, a term now used to mean either a brand mention inside an AI answer or an AI crawler visiting your site, with no standard definition yet.
- And branded search trends in Google Search Console, which she’s clear are directional at best. “It may not even be 70% accurate for you,” she says, “but it is a good directional signal.” Read the trend, not the number.
Underneath both sits a problem she says SEOs created for themselves: years of investing in top-of-funnel content because it drove traffic, leaving middle-funnel content (comparison, consideration) thin almost everywhere. That gap is exactly where an AI answers a question your own site never addressed.
Local AI visibility trackers can help you see how you’re being mentioned when tools like Search Console or Analytics fall down.
Dana’s practical method for finding the content you’re missing
Dana’s method for finding that gap is a repeatable process.
- Export around 90 days of Google Search Console query data as a CSV
- Feed it into Claude alongside your keyword research.
- Connect any SEO tool that has an MCP (she name drops SE Ranking and BrightLocal’s MCP, then mentions DataForSEO and SEMrush as other options), so ranking data is in the same conversation.
- Ask it to bucket the queries by funnel stage and flag what’s underserved.
- Then, ask it to prioritize against your real available time rather than handing you an overwhelming list: “I’m a team of me, I can write for a half day on Fridays, where should I start?”
“I remember going through lists of keywords in Excel and just manually sorting, doing a VLOOKUP of the different things … No, it’s so much faster now.”
Once you’ve found the gap, match how people actually phrase questions to AI, not just head keywords. Nobody types “budgeting software” into an LLM; they type “I’m a small business with two million dollars in revenue in British Columbia, looking for budgeting software, and I have an accountant I send my books to.”
An AI-trained reporting assistant gets you past the blank page
Dana has trained a Claude project on everything she’s ever said about building reports in Looker Studio. She fed it her formulas, her preferences, and what she does and doesn’t want in a dashboard. Now you give it a client brief or a draft post, and it returns a markdown spec with the audience’s questions, a proposed dashboard structure, the microcopy for each chart, even the “about” page she always starts with.
For example, she was writing a post for Analytics Playbook asking, “are Google Search Console impressions real, and how do you spot it if they’re not?”
She fed the draft in and got back a full dashboard spec built around that question, including a bar chart, a heat map, and a calculated field she ended up rejecting.
“I don’t think I’m actually going to need this. I think I can pull the data out of GA4 directly.”
It may seem like something small, but this is where having an actual expert build your AI can make a real difference. Rather than simply accepting what was created as a given, she questioned it and stopped AI busywork from being created.
“It saves us, I would say, a few hours at least when we’re creating a dashboard from scratch.”
It isn’t a finished product; there’s still no direct connection between Claude and Looker Studio, so building the actual charts is manual. What AI has changed is the thinking time you need to get this started.
The context she feeds it matters as much as the tool itself. She’s explicit that AI without a business context will get basic things wrong.
For example, Dana was asking Claude whether sessions were up or down without giving it seasonality context, before catching herself.
“No, well, yeah, it’s Christmas… that’s not [a major time] for a B2B business, obviously nothing happens at Christmas, nothing happens in the summer.”
So what sort of thing can you feed into to get this context?
Client call transcripts, Notion records of major decisions, and a short onboarding survey (how this client is evaluated, what chart last helped them make a decision) should all go in.
Without that layer of context, the output is generic.
Why the giant dashboard is dying
Dana is openly against what she calls “Star Trek style” dashboards, the three-monitors-wide kind packed with every metric available. Her view of where reporting is headed is:
- 3 to 6 metrics people actually check regularly,
- Plus a self-serve layer where clients can ask their own questions, with her firm’s judgment already built in.
And she highlights that the expert judgment on what to include really matters. She wants the system to refuse to report a metric like engagement rate, even if a client asks for it directly, because it’s meaningless, and to explicitly explain why.
It’s an early idea; she’s exploring whether an MCP can deliver this for clients directly, working alongside Britney Muller on whether it’s technically possible, and deliberately avoiding shared Claude projects across organizations for now.
“There’s nothing worse than making a dashboard, and nobody uses it.”
What clients actually ask their own data
One of Dana’s more unusual habits is that she wants to see the literal prompts clients type into ChatGPT or Claude about their own numbers. She even asks them to send screenshots. Her reasoning is that what a client asks, and how they phrase it, tells her both what her reporting is missing and how sophisticated that client actually is with the data.
It’s the same logic as a “how did you hear about us” form. People answer with whatever’s most recent, not the full picture, so watching the actual questions catches what a client never thought to mention.
Getting started, whatever your executive function looks like
Dana mentions, almost in passing, that she has ADHD and that AI’s biggest personal value to her is executive functioning. She sees it as something to tell her where to start. It’s the same idea the whole conversation keeps returning to. AI’s job isn’t to finish the work. It’s to get you past the moment where you’re staring at nothing, so you can get on with the part that actually needs your judgment.
