AI-Powered, Human Approved Agency Insights


More How, Less Hype

Hi Reader,

If you haven't been living under a rock for the past few years, your inboxes, feeds, and networking events are filled with people breathlessly talking about the potential of AI.

Pilots. Proof of concepts. It's all about Potential.

Instead of more hype, we're going to show you every place we use AI in our Small Agency Assessment (currently on a Black Friday special), why we're using it, and how we compensate for its shortcomings.

1. Call Analysis

We use Fathom throughout the Assessment to transcribe calls, generate better-than-Zoom-generated summaries, and ask follow-up questions.

One feature we love is how with two clicks you can reframe the call summary into a number of popular sales frameworks (BANT, MEDDPICC, SPICED) which helps uncover where agencies are having the biggest struggles.

After the Assessment intake calls, we copy the AI-generated summary into a Google Doc and edit it extensively based on our personal observations that weren't caught by the AI.

For example, AI doesn't catch body language or nuanced tone of voice, but we've seen what shame, avoidance, and embarassment looks like. Those emotional cues tell a much deeper story that helps us calibrate our recommendations, such as leaning into confidence-building versus risk-taking activities.

Those edited Google Doc summaries then go into-

2. Deeper Analysis

Call summaries, transcripts, and our secret sauce of agency best-practices gets organized in NotebookLM, a tool that helps us connect dots and ask better questions. NotebookLM is better than ChatGPT or your AI tool of choice because it limits the scope to the materials we upload.

Great advice for a mega-agency like GroupM is terrible advice for a three-person SEO agency: NotebookLM helps us stay on the straight and narrow.

We use NotebookLM to create a draft Assessment Report. It's usually a little too generic and far too ambitious: if you follow its raw advice, you'll be investing time into too many activities to make any noticeable impact.

3. Polishing the Analysis

Once we've tailored our draft Assessment Report, we run it through a custom-built Google Gemini Gem. This Gem is coded to behave like a small agency owner, foibles and all.

This double-check helps us make sure we're on-target with the recommendations, rephrase certain points, and add nuanced "Why" explanations into the Report.

Unfortunately, just like NotebookLM, Gemini also tends to push recommendations to the middle of the road and is just as overambitious. Massaging the final Assessment Report is best done by human hands.

4. Making The Analysis More Accessible

After all of this, we load the final Assessment Report into NotebookLM and then generate an audio overview. This helps Assessment clients consume and understand the report in an engaging manner.

We have to add several layers of direction into that audio overview so it focuses on the recommendations, otherwise you are rolling the dice that it produces something useful to an agency owner.


Clear the Noise With A Small Agency Assessment

If you want to start 2025 with clarity on your biggest opportunities and risks, our Small Agency Assessment comes with a $750 coupon for any of our standard trainings in 2025. This offer only lasts until January 1st, so book an assessment today!

Dan Newman

I help organizations build AI fluency and governance that actually changes behavior — not the kind that lives as a PDF on a Notion page. 19 years onboarding humans to strange new places (startups, scaling tech, enterprise agencies like GroupM and WPP) gave me a head start when AI showed up as just another strange new place. The TL;DR is my biweekly newsletter for leaders thinking through what AI means for their people.

Read more from Dan Newman

A bi-weekly roundup of pushing boundaries October 31 - November 13 Do What Feels Good What is the first software you reach for on Monday morning? Your email? Your calendar? A project management platform, like Asana or Trello? Or maybe a dashboard in Hubspot/Salesforce? I've noticed in the past three months that for me, my Monday software first touch is increasingly Claude. And not only is it becoming the first tool I load up in the morning while eating yogurt, but I've even been shifting...

Storybook-style illustration of a cozy suburban front door at dusk, standing slightly ajar with warm golden light spilling onto the porch. A doormat reads "COME IN, WE'RE HELPFUL." Peeking from the doorway is a small gremlin with wrinkled greenish-brown s

A bi-weekly roundup of big tech trying to snooker our data, like they always do October 17 - October 30 We Shall Fight On The Beaches (against Muse, Pulse, and Duet) Yesterday I made pulled pork in a homemade pepper-jelly over rice, topped with pickled onions. It was a recipe that Toddlywonks, my Muse AI, had generated after I had scanned in my pantry and the frozen meat. TL;DR- it was incredible. But what was not incredible was how annoying Toddlywonks was during the process. This was a sous...

A chart showing Job title searches (peaked in April 2023) and Skill listed in postings (Peaked 2024)

A bi-weekly roundup of depreciating skills October 3 - October 16 Stop trying to make 'Prompt Engineering' happen. It's not going to happen! I remember when I was advised to list "Proficient in Microsoft Word" on a resumé. If I did that today, I'd probably look like a doorknob. It's pretty obvious that some skills have become so fundamental in today's workforce that calling them out would work against you, but when will "Proficient in AI" become the new Microsoft Word? Is that day already...