In this blog post I'm going to walk you through what actually happened when I went all in on AI in my marketing business in 2024. The version with the failures included. The numbers, the workflows that worked, the workflows that didn't, and the things I'd do differently if I were starting again.
Most "I built my business with AI" content is performative. It's written by people who started yesterday, who haven't shipped anything to a paying client, or who are selling a course about the thing they're describing. This is different. I've been a marketing consultant for twenty-one years. I went all in on AI in 2024 because I had to — five brutal personal years had collapsed my pipeline. By early 2025 I was rebuilding from a much smaller base. Fourteen months later, here's the report.
By the end of this blog you'll know exactly which AI workflows produced revenue, which produced nothing, what I spent, what I'd spend again, and what the honest learning curve looks like for a non-technical marketer going deep on AI.
TL;DR
What worked: - Custom AI-built WordPress publishing pipeline (saved roughly 400 hours/year) - AI-driven SEO recovery on the site (sitemap grew from 904 to 1,300+ URLs in 3 weeks) - AI for sales call prep, follow-up, proposal drafting (cut sales cycle by 60%) - AI-assisted content production for the newsletter (15k subs, 70% open rate maintained) - Building tools, calculators, and lead magnets I would have outsourced before
What didn't: - AI for "thought leadership" at scale (engagement collapsed) - AI-generated outreach (response rate dropped, killed personal brand) - AI-only content creation (Google noticed) - "AI marketing strategy" tooling (produced strategy decks nobody used) - Specialised vertical AI tools (too narrow to justify subscription)
Net: revenue rebuilding ahead of plan, costs down, work I love increased.
The starting point
January 2025 starting position: - 21 years experience as a marketing consultant - Established personal brand (15k newsletter subs, decent inbound) - Pipeline collapsed after 5 years of personal upheaval - Almost no AI experience beyond ChatGPT-as-search-engine - Non-technical (no coding background)
Decision: go all in on AI as the new positioning, the new operating system, and the new service offering. Not "I help businesses with AI" the way every former marketer was suddenly claiming. AI as the thing that would let me deliver more value at smaller margins with smaller teams.
The bet: if I could actually use AI well enough that my own work improved measurably, I could sell that improvement to others. If I couldn't, the whole positioning would collapse and I'd need a different plan.
Month 1-3: The infrastructure investment
Most "I rebuilt with AI" stories skip the infrastructure phase because it's not glamorous. It's also the part that determines whether everything that follows works.
What I built in the first three months:
Custom WordPress publishing pipeline. Built using Claude Code in a series of weekend sessions. Takes a finished blog draft, generates SEO meta, creates a featured image, populates Yoast fields, uploads as draft, sets internal links. The version I'm still using is at roughly 4,000 lines of Python plus integrations with Bing Webmaster Tools, GA4, and Search Console.
Time spent building: roughly 60 hours over 6 weekends. Time saved since: about 400 hours/year, ongoing.
Newsletter automation. Weekly Kit broadcast pipeline. Doc → draft email → review → send. The automation never sends without me — hard-coded constraint.
Time spent: 12 hours. Time saved: 90 minutes per week (78 hours/year).
Inbox triage system. Reads my Gmail, classifies messages, drafts responses, queues for review. Took multiple iterations to get response quality right.
Time spent: 30 hours. Time saved: about 4 hours/week of decision overhead (200 hours/year).
Sales call infrastructure. Granola for call transcription, Claude for prep briefs from prospect LinkedIn + website, Claude for follow-up drafts.
Time spent: 8 hours. Time saved: 90 minutes per sales call across 4-8 calls/week. Roughly 250 hours/year.
Total infrastructure investment: about 110 hours. Total annual time recovered: about 928 hours. Payback: 5.7 weeks.
The infrastructure wasn't the visible work. But without it, none of the visible work would have been possible.
Month 4-6: The content reset
After infrastructure came the content reset.
The mistake I made in month 4: I tried AI-only content for the blog. Volume tripled. Quality looked fine on read-through. Six weeks later, Google's helpful content update penalised the site. Organic traffic dropped roughly 35%.
The fix: hybrid model. I write the outline, the opening, the contrarian take, and the close. AI helps with research, structure validation, FAQ extraction, and meta data. The 80/20 split is roughly 80% human voice, 20% AI scaffolding.
Three months of recovery work later, traffic was back above baseline and the new posts were ranking. The lesson: AI as a multiplier on judgement works. AI as a substitute for judgement gets flagged.
What I'd do differently: skip the AI-only phase entirely. The temptation to scale content is real. Resist it.
Month 7-9: The service redesign
Before AI, my services were: marketing consulting, fractional CMO, retained advisory. Hourly or monthly billing. Standard.
By month 7, my own AI workflows were producing 3-5x output at similar quality. If I kept billing the same way, AI would cannibalise my own revenue. I had to redesign.
The new service set:
What changed: pricing moved from time-based to outcome-based. AI efficiency now increases margin instead of cannibalising revenue. The service set is more selective — I do fewer things and charge more for them.
The hardest part: convincing past clients that the new pricing reflected better outcomes, not just better margin for me. About 60% transitioned to the new model. The other 40% either left or stayed on legacy pricing through their existing contracts.
Month 10-12: The positioning shift
The first nine months were operational. By month 10, the positioning question couldn't be avoided.
The temptation: become "AI marketing expert" alongside every other former marketer. The problem: that category was already saturated with people whose AI experience was 6 months.
The choice I made: position as the "AI consultant for the non-technical." Twenty-one years of marketing, deeply operational AI experience, no engineering background. The audience: founders and marketing leaders who feel locked out of AI because every guide assumes Python.
Specific positioning calls: - Kept "marketing consultant" in the title (don't abandon two decades of equity) - Added "AI implementation" as the modifier - Wrote publicly about my own failures with AI (the AI-only content disaster) - Stopped writing about tools, started writing about workflows and decisions
Result by month 12: inbound enquiries shifted from "can you help with our marketing" to "can you help us figure out AI." Better fit, higher prices, faster closes.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Month 13-14: The sales motion
The current state.
Sales motion: - Inbound through the blog, newsletter, LinkedIn - 4-8 discovery calls per week - Discovery → audit (paid) → project (if fit) - Average sales cycle: 18 days (was 45 days before)
AI's role in sales: - Granola transcribes every call - Claude prepares the brief from prospect's LinkedIn + website (saves 45 min per call) - Claude drafts the follow-up email within 30 minutes of the call ending - Claude drafts the proposal from the discovery transcript (saves 3-4 hours per proposal) - I review and edit everything before it goes out
What I do not delegate to AI: - Reading the prospect on the call (tone, hesitation, what they're not saying) - The yes/no fit decision - The pricing conversation - Anything that requires reading the room
The financial picture
Revenue by quarter (rebased to January 2025 baseline = 100):
- Q1 2025: 100 (baseline)
- Q2 2025: 145
- Q3 2025: 220
- Q4 2025: 285
- Q1 2026: 340
Costs: - Infrastructure maintenance: 2-4 hours/week of my time
Notable: zero engineering hire. Everything I've built has been built by me, in Claude Code, on weekends. This is the part I want non-technical founders to take seriously. If a non-technical marketing consultant can build production AI workflows, you can too.
What worked (extended)
The custom publishing pipeline
This was the single highest-ROI thing I built. Once it existed, content production stopped being a bottleneck. Featured image generation, SEO metadata, Yoast field population, draft uploads, internal linking — all automated. I spend my time on the writing and editorial calls, not on the infrastructure of publishing.
If I were starting again I'd build this first, not third. Get the publishing flywheel working before anything else.
AI for SEO recovery
I had 21 years of marketing experience but limited deep SEO experience. When my site's organic traffic collapsed, I used AI heavily to diagnose. Schema markup additions, sitemap regeneration, internal linking improvements, FAQ schema, HowTo schema, redirect mapping. AI made me competent at SEO operations far faster than reading would have.
Specific result: sitemap grew from 904 to 1,300+ URLs in three weeks. Stuck-out-of-sitemap posts unstuck via a "re-save trick" I found through AI-assisted experimentation.
AI for sales call prep and follow-up
The single highest-leverage sales improvement. I now spend 5 minutes on call prep where I used to spend 50. I now spend 15 minutes on follow-up where I used to spend 90. The quality of both has improved, not declined, because the AI catches things I would have missed (prospect's recent activity, competitor mentions, specific terminology they use).
Newsletter production
The newsletter has held its 70% open rate through the AI integration. Writing remains mine. AI helps with research, structure validation, subject line testing, and segmentation. The newsletter is more frequent now (was every 2 weeks, now weekly) without a quality drop.
Tools, calculators, and lead magnets
This is where AI's leverage as an enabler matters most. Before AI, I would have hired a developer to build any interactive tool. Now I build them myself in a weekend. The Lilach Bullock site now has multiple interactive tools (a cost calculator, an AI readiness audit, an ROI estimator) that all generate qualified inbound leads. None of them required external developer cost.
What didn't work (extended)
AI for thought leadership at scale
I tried scaling my LinkedIn content with AI assistance. Three posts a day instead of one. Engagement dropped 80%. Comment quality dropped to nothing. The audience knew. Cut back to one well-considered post per day, engagement recovered.
Lesson: distribution channels reward signal over volume. AI is great for first drafts of single posts. It is terrible for replacing the editorial filter that makes you worth following.
AI-generated outreach
Tested for two weeks. Response rate dropped from about 12% to 3%. People could tell. I got two negative replies from people I respect saying my recent emails felt generic. Stopped the experiment immediately.
What works instead: AI for prospect research (background, recent activity, likely interests), human for the actual outreach email.
AI-only content
Covered above. Don't do it. AI-assisted, human-led. The line matters.
AI strategy tools
Specialised vertical AI tools
I subscribed to several specialised AI tools in different verticals (legal, healthcare, finance) thinking I'd resell their outputs in consulting projects. The tools were thin wrappers on the general-purpose LLMs with worse UX and higher prices. Cancelled all of them within 60 days.
What I'd do differently
Three things, in order of impact:
1. Build the infrastructure first. I spent month 1-3 doing client work while learning AI on the side. I should have spent the first 6 weeks heads-down on infrastructure. The compounding returns from doing this earlier would have been enormous.
2. Skip the AI-only content phase entirely. The traffic recovery cost me 3 months. I should have known better. Don't repeat the mistake.
3. Redesign pricing before scaling AI. I scaled AI in my delivery while still billing hourly. That cannibalised my own revenue for two quarters. If you're billing time-based, redesign pricing first or AI will eat your margin.
What's next
Year 15-24 plan (because I'm not stopping):
- Build out the AI implementation consulting practice (current focus)
- Document the full operational playbook publicly (this blog plus newsletter)
- Productise the audit (currently 1:1, eventually scaled)
- Continue building tools I'd otherwise outsource
- Maintain the 80/20 human/AI ratio in all visible content
The thesis remains: AI is a tool. Business is the game. I've been playing the game for 21 years. AI lets me play it harder, not differently.
Frequently asked questions
How long does it really take a non-technical person to get good at this? Honest answer: 6-12 months of consistent daily use. There's no shortcut. The people who claim to have transformed in 30 days have not actually transformed.
Did you hire any engineers? No. Everything built has been built by me in Claude Code on weekends. I have a strong technical understanding from 20+ years of running marketing-adjacent technical projects, but no code-writing background.
What was the hardest part? The AI-only content failure. Watching traffic drop 35% and knowing it was my fault took real discipline to fix calmly instead of panicking.
What was the best investment? The infrastructure work in months 1-3. Highest ROI of anything I've done.
Would you recommend going all in like this? Only if your existing business is small enough to redesign and your appetite for short-term revenue volatility is high. If you have 20 employees and stable revenue, do this incrementally.
Are you ever going to stop and run a "normal" consultancy? No. The fun has come back. The work is more interesting. Revenue is rebuilding. The path forward is more AI, not less.
Want to build this for yourself?
If the audit concludes "you don't need an AI consultant, fix X first," that's the conclusion. You walk away with the diagnosis and no further pitch.
I'm Lilach Bullock. I've been a marketing consultant for twenty-one years. I went all in on AI in 2024. I work with founders and marketing leaders who want AI to actually move their numbers, not just their tool stack.
Related from this week
- /ai-implementation-roadmap/ — the 90-day path I actually use
- /ai-for-marketing-agencies/ — AI for agency operators specifically
- /ai-for-seo-without-penalty/ — the AI-safe SEO playbook
