[The following article was derived from an unstructured monologue that was automatically recorded from my new Limitless Pendant while I was getting the kids ready for the day (no AirPods, no phone). I just spoke loud and then popped it into both ChatGPT's o1 Pro model and Gemini 2.5 Pro, iterating back and forth asking them to critique each other's drafts until I had something I was happy with (at least enough to spend another 10 minutes putting finishing touches on tagging vendors etc.). I'll continue to do more of these experiments with nascent ambient compute and share the results on LinkedIn.]
Something I need to get done today: creating a targeted growth campaign for AI services in the legal sector-especially for smaller law firms. As I talked through the steps, I realized it's a perfect example of where we are now with AI tools, where we could be soon, and what it might look like when everything is frictionless in the future.
My Current Reality: So Many Steps, So Many Clicks
I'm juggling a handful of AI and automation tools-everything from email-sending assistants (via Shortwave | We're Hiring!) to semi-automated product development pipelines (e.g., Fireflies.ai transcripts captured from customer discovery interview -> Claude Code).
Right now (esp. in the context of the new campaign I want to launch), if I want to craft a new landing page or do in-depth research, I have to:
Retrieve transcripts from my meeting recordings (I store these in Fireflies, among other places).
Search across multiple accounts, tabs, or scripts to gather the insights I need.
Spin up code (local React app dev via CC and GitHub or via a cloud dev environment / "vibe coding" product like Lovable) to update my site and deploy with Netlify.
Create email drip campaigns, SMS blasts, scheduling workflows, and voice-driven intake and qualification with a suite of marketing tools: e.g., Calendly, HubSpot, Vapi.
Research top-of-funnel targets and find relevant podcasts or communities (ChatGPT / Gemini Deep Research -> analysis of past transcripts by pulling from the Listen Notes, Inc. API to ensure I can contribute to their platform and not repeat whatever past guests have said) to pitch directly or via a broker like PodMatch .
Manually integrate everything back into a coherent and cohesive funnel-landing pages, sign-up forms, analytics, you name it.
It all works. But it's a lot of friction and definitely not the super-smooth, "AI-does-it-all" dream we keep hearing about.
Why Law Firms?
I've already helped several law firms-from criminal defense and personal injury to HOA specialists-use AI in their operations. Each of those engagements generated a ton of potentially valuable data: transcripts, meeting recordings, tool configurations, results, and so on. This vast collection of interaction data-our "digital exhaust"-becomes the essential raw material that future AI agents will mine for context, personalization, and proactive insights. Having a track record with four (soon to be more) law firm clients means there's enough credibility and "compelling facts" to do a real campaign.
Why focus on AI for lawyers?
Many are looking for an edge-efficiency, speed, better client service as AI drives down the cost of reading, writing, and reasoning tasks.
It's a conservative industry, so showcasing real-world success stories matters.
There's a massive opportunity to automate tasks like drafting hyper-personalized demand letters, summarizing depositions, and even reviewing evidence with a fine-tooth comb + suggesting cross-examination questions.
The Campaign Strategy
I want to create a hyper-targeted campaign that might include:
A Landing Page specifically for law firms interested in AI, with copy and content derived from the good work I've already done.
A signup form for a Webinar or "live AI magic show" that demonstrates these tools in action (60-90 minutes, Q&A, etc.); I do these all the time.
An Email Drip or SMS follow-up sequence (probably in HubSpot, though I may check other platforms).
Podcast Outreach to appear on niche lawyer-focused shows (via Podmatch or direct outreach) for top-of-top-of-funnel.
Deep Research Reports (e.g., "What personal-injury firms in New Mexico are doing with AI right now," "what precedent has been set in Rhode Island for AI use in DUI discovery") as lead magnets to attract interest and market the webinar via a complementary drip campaign.
Why Do All This?
Thought Leadership: Doing it live, showcasing real client wins, and clarifying exactly how AI can transform a law firm's practice.
Efficiency: A big webinar means I can teach multiple firms at once, saving me from repeating the same demos.
SEO & Content: Blog posts, FAQ pages for lawyers, and in-depth tool comparisons help attract inbound interest and "capitalize" on my unique data asset, which grows with every client engagement.
The Friction: Where AI Falls Short (for Now)
Despite having good AI tools, I still have to do a lot of manual orchestration. For example, if I want a transcript-based report on "podcasts that suit my AI-for-lawyers angle," I'll:
Pull transcripts from open Chrome tabs via the open-source Raycast script I wrote.
Feed that into a reasoning model + Deep Research to scour the internet for salient, niche podcasts.
Review the summarized results and decide which three to five podcasts to target.
Craft a pitch
Double-check everything (and probably at every incremental step) for accuracy.
This is still "the future," but it involves more busywork than I'd like; it can all be accomplished in 1-2 deep work sessions, but there is a ton of clicking and context switching, which can be draining. In my ideal world, my wearable AI (like my Limitless Pendant that's always listening, at least by default) would just parse my morning monologue, spin up an agent, and handle these steps invisibly-only pausing to confirm details or alert me when a big decision or next-level budget is required; my intent is clear enough that the system can take things as far as it's able to before it hits an expensive pocket of uncertainty that demands my own "executive" attention.
A Vision of "Ambient" AI
Imagine hooking up multiple MCP (Model Context Protocol) servers or other orchestrator systems (Google's new A2A Protocol might be an interesting complement) that let different specialized agents collaborate:
Agent 0 (powered by the smartest, most expensive model) comes up with a plan
Agent 1 retrieves meeting transcripts automatically.
Agent 2 handles website and landing-page updates ( Cognition 's Devon?)
Agent 3 scans Podmatch and other directories for relevant podcasts, crafts an intro email or pitch, and queues it for my review.
Agent 4 sets up the webinar funnel, integrates analytics (Google Tag Manager, etc.), and schedules test runs.
Everything happens behind the scenes. The AI pin just listens to me brainstorming, identifies tasks, and delegates to specialized agents. Crucially, these systems wouldn't just reactively execute commands; they would proactively bootstrap new opportunities (and formalize goals or acceptance criteria that drive the activities of all of the other agents à la a disciplined OKR process), draft initial plans, or suggest strategies by continuously analyzing patterns and latent needs within our digital exhaust.
When done well, this would feel like having a top-tier enterprise department on standby 24/7, all triggered by a casual voice note-or even working proactively to accelerate progress toward your quarterly objective, all driven by the proprietary, salient, high-fidelity digital exhaust you (and your collaborators) generate daily.
The Long Game: Everyone Gets a Mini Corporation
Eventually, I see a world where each person (or small business) has a "swarm" of AI agents acting like an invisible workforce-orchestrating tasks, collecting data, and running experiments. You feed it the bigger goals-like "Expand my AI consulting into 10 new law firms"-and the system works out how to get there: from budgeting to scheduling to copywriting w/ limited tugging on your scarce attention.
We're not fully there yet (and not even close). But by capturing not just the product of your expert/entrepreneurial labor, but also the process, at least latently in timestamped, thoughtfully indexed primary sources generated through nascent ambient computing sensors-every meeting, every transcript, every piece of feedback-you can build the data foundation that tomorrow's agent swarms will need to help you GSD 10x faster.
Conclusion: It's Here, But It's Not
So, that's where my head is right now: pushing forward on a lawyer-focused AI campaign while juggling a bunch of separate tools. It works, but it's not seamless. Give it five, ten, or maybe fifteen years, and I suspect we'll look back on all these clicks and manual steps the same way we look at dial-up internet today.
In the meantime, if you're eyeing the AI frontier-especially in a niche industry-my advice is simple:
Capture Everything - Transcripts, steps, successes, fails. This data, your digital exhaust, will feed future automations.
Define Specific Goals - Pick a niche (like small law firms) and go deep (deeper than anyone else ever has or could, which is increasingly feasible as the cost of machine intelligence continues to fall like a rock) with a clear offering, derived from your initial, successful, manual attempts at solving new problems in new ways.
Use AI Where It Shines - Summaries, outlines, quick code fixes-and then manually fine-tune the last mile.
Stay Nimble - The next big shift in AI could be tomorrow. Remain ready to pivot, and don't get overly attached / over-engineer your current processes & AI tech stack.
We can see how powerful it'll be when these friction points are smoothed out. For now, though, I'm embracing the scrappiness-clicks, scripts, and the occasional under-the-bed toddler rescue (which apparently was woven into the original transcript for this piece)-and forging ahead.