AI content marketing for teams of one: 90-day plan

Tahi GichigiThu Jul 02 2026 · 10 min read

If you are a solo marketer (or a founder doing marketing on the side), “use AI for content” usually fails for one reason: it still needs you to drive. Prompts, rewrites, chasing reviews, copying into the CMS, tracking results.

This plan assumes a realistic goal: one searchable post per week that answers a buyer question and supports one conversion.

What “AI content marketing” means when you are a team of one

The outcome: weekly, searchable posts that earn trust

Stop chasing “more content”. Set one outcome you can execute:

• 1 post per week
• 1 search intent per post (one query family, not a pile of keywords)
• 1 primary CTA (demo, trial, pricing call, newsletter)
• Proof-led writing (screenshots, numbers, named tools, benchmarks)

AI helps with research, drafting, on-page SEO, and analysis. The risk is using it as a faster typewriter and producing generic posts that do not convert.

Split the work into five jobs (so nothing falls through)

If “AI content marketing” is a blur, it stays stuck. Break it into jobs with clear outputs:

1. Positioning capture: one page that nails who you sell to, what you promise, and what you refuse to say.
2. Topic system: a repeatable way to pick topics based on demand and fit.
3. Draft production: research, outline, draft, on-page SEO.
4. Approvals: who checks what, by when, with default actions.
5. Publishing and measurement: ship on schedule, track performance, iterate.

This is consistent with mainstream guidance on AI in marketing and AI-assisted content workflows, but it adds the part most articles skip: defaults and constraints for a team of one (Salesforce: https://www.salesforce.com/marketing/what-is-content-marketing/ai-content-marketing/, IBM: https://www.ibm.com/think/topics/ai-in-marketing, Airtable: https://www.airtable.com/articles/ai-content-marketing).

Three non-negotiables: one audience, one promise, one conversion

Before tools, set constraints:

• One audience for 90 days (example: “RevOps leaders at B2B SaaS, 20 to 200 employees, HubSpot + Salesforce”).
• One product promise you can defend (example: “Cut onboarding time by 30% without hiring”).
• One primary conversion per post, fixed for a month at a time.

Constraints prevent polite, generic content that attracts the wrong clicks.

Days 1 to 7: capture positioning and voice once so AI stops guessing

Write a one-page positioning brief (60 minutes, not a manifesto)

If it takes longer than an hour, you are writing something nobody will use.

Include:

• ICP: role, company size, stack, buying trigger
• Core pain: what breaks, what it costs, who gets blamed
• Why now: what changed (market, regulation, budgets, tech shift)
• Proof points: 3 to 5 facts (case study metric, customer names you can use, credible benchmarks)
• Top objections: 5 blunt objections with blunt answers
• Terminology: words you will use, and words you will not (“forbidden phrases”)

This is what stops AI from rewriting you into generic SaaS-speak.

Build a voice pack (rules plus examples)

AI matches tone when you give it anchors.

Your voice pack:

• Tone rules: direct, evidence-led, no hype
• Formatting rules: short paragraphs, specific headings, bullets over long prose
• Claim rules: no sweeping claims without proof or a source
• Examples: 5 to 10 samples (homepage copy, sales emails, support docs, best posts)
• Do-not list: banned openings (“In today’s world”), vague claims (“game-changing”), empty modifiers (“seamless”, “robust”)

Pick 2 to 3 posting lanes tied to buyer searches

Avoid broad themes (“AI”, “productivity”). Choose lanes you can own and that map to buying intent.

Example lanes for a B2B analytics product:

• Implementation and operations (setup, governance, workflows)
• Reporting and metrics (definitions, templates, best practice)
• Tool selection and comparisons (build vs buy, vendor comparisons)

Rule: every post must fit a lane.

Days 8 to 21: install a topic system that does not rely on inspiration

Run a lightweight content gap audit by intent cluster

You do not need an enterprise audit. You need a backlog you can publish from.

Steps:

1. Export your posts into a sheet: URL, title, target query (if known), last updated.
2. Pick 3 to 5 competitors (direct competitors plus “content competitors” that rank).
3. Pull their top pages and keywords using Ahrefs or Semrush (or manual SERP review if you have to).
4. Group by intent cluster, not keyword variants.

Common intent clusters:

• “How do I …”
• “What is …”
• “Best …”
• “X vs Y”
• “Alternatives to …”

AI can summarise SERPs and competitor angles. Your job is to reject clusters that do not match your ICP and product. That is where scaled AI content often goes wrong (Salesforce link above).

Use one brief template so you can create 30 briefs quickly

Copy and paste this 30 times.

Content brief template

• Target cluster:
• Primary query:
• Search intent (what they are trying to decide or do):
• Job-to-be-done: “When I am…, I want to…, so I can…”
• Angle (your specific take tied to positioning):
• Must-cover sections (bullets):
• Proof required (screenshots, data, named tools):
• Internal links (3 to 5 URLs):
• Primary CTA:
• Secondary CTA (optional, only if it does not distract):

If you cannot fill the “angle” and “proof required” sections, do not write the post.

Build a 30-post backlog using three buckets

Fill 10 posts per bucket:

1. BOFU comparisons

   • “X vs Y”
   • “Best tools for [job]”
   • “Alternatives to [competitor]”
   • “Pricing and implementation for [category]”

2. How-to problem solving

   • “How to [do job] without [constraint]”
   • “Checklist: [process]”
   • “Template: [document]”
   • “Troubleshooting: [failure mode]”

3. Category education tied to your product

   • “What is [category], and when is it worth it?”
   • “Common mistakes in [category]”
   • “Metrics that matter for [job]”
   • “How to evaluate [category] for [ICP]”

Cadence rule: rotate BOFU → how-to → education.

Days 22 to 35: design a minimum-effort workflow (draft to publish) with approvals

Pick a workflow level and turn it into policy

Match your risk profile:

• No review: publish weekly without approvals.
• Light review: approve outline only.
• Full review: approve final draft.

Most teams of one should default to light review. It prevents major misalignment without killing throughput (Airtable link above).

Set SLAs and default actions so review cannot kill cadence

Publishing dies because “when you have time” becomes never.

Set:

• SLA: 48 hours for feedback.
• Default: if no feedback in 48 hours, publish.
• Escalation: one reminder at 36 hours, then proceed.

If you need legal/compliance:

• Use a checklist (claims, customer names, regulated terms, disclaimers).
• Route only posts that trigger the checklist.
• Keep legal review to final draft only.

Standardise the boring assets once

Standardise these so every post can ship without handcrafting:

• Featured image template: one style, fixed dimensions, keyword in filename and alt text.
• CTA blocks: 2 to 3 reusable blocks (demo, pricing, newsletter).
• Metadata rules: title tag pattern, meta description pattern, FAQ schema where relevant.
• Internal linking checklist: product page, one integration page, one “start here” post, two related posts.

AI is useful here in a practical way: drafting metadata variants, suggesting internal links, and producing consistent structured sections (IBM link above).

Days 36 to 60: publish weekly with a cadence that is hard to break

Commit to one publish day and run a weekly rhythm

Pick one day you publish. Protect it like a customer meeting.

A workable rhythm:

• Monday: finalise brief, list proof sources you need
• Tuesday: draft
• Wednesday: review (outline-only or final)
• Thursday: publish and distribute
• Friday: log results and queue repurposing

The goal is throughput.

Use single-input reviews (fact check, product accuracy, risk)

Give reviewers one job: correct facts, claims, and product details.

Review is not for:

• rewriting for “tone”
• swapping headings for preference
• adding pet ideas

If style is off, fix the voice pack, not the draft.

Create a repurposing loop that pays back your time

Every post produces:

• 3 LinkedIn posts (insight, checklist snippet, contrarian takeaway)
• 1 email snippet (100 to 150 words, one link)
• 1 sales enablement note (one paragraph: “send this when a prospect says X”)

If you cannot repurpose it, it is usually too generic.

Days 61 to 90: measure what matters and let performance steer the next month

Track a small scorecard that maps to pipeline

Track weekly, review monthly:

• Impressions and clicks (Google Search Console)
• Positions for 1 to 3 target queries per post
• Assisted conversions (GA4, HubSpot, or CRM attribution)
• Sales usage (links shared in emails, call notes, enablement)

AI can help spot patterns across clusters and recommend what to double down on, which is a core use case in AI-driven marketing systems (IBM link above).

Use decision rules so you do not debate when you are busy

Rules you follow automatically:

• Refresh posts sitting on page 2 (positions 11 to 20): tighten intro, add missing intent sections, improve internal links, add proof.
• Double down on clusters that drive qualified demos: write adjacent posts in the cluster, add comparisons, tighten CTAs.
• Prune vanity topics: if a post has traffic but no assisted conversions after 60 days, reposition the angle and CTA or stop that cluster.

Run a monthly retrofit cycle (half a day)

Retrofit the last 8 to 12 posts:

• Add internal links from older posts into recent winners
• Update stats and examples
• Rewrite intros so payoff is in the first 3 lines
• Add missing intent sections (“Who this is for”, “Costs”, “Implementation pitfalls”)
• Recheck CTAs (one primary conversion)

Publishing is only half the system.

Common failure modes (and fixes you can apply in a day)

Failure mode: AI output looks generic

Generic output is usually missing constraints.

Fix:

• Strengthen the one-page positioning brief.
• Add a forbidden phrases list.
• Require proof sources in every brief (screenshots, numbers, named tools).
• Reduce lanes to 2 to 3 you can support.

Failure mode: publishing stalls at review

Fix:

• Move to outline-only approvals unless you are regulated.
• Enforce 48-hour SLAs with default publish.
• Use a risk checklist instead of reviewing every post.

Failure mode: traffic grows but pipeline does not

Fix:

• Increase BOFU coverage (comparisons, alternatives, “best tools”).
• Put one clear CTA near the top and one at the end.
• Add proof-led sections (customer examples, benchmarks, implementation detail).

Education content without proof does not build trust.

How self-driving content changes the 90-day plan

Most “AI content tools” still need you to drive: prompts, edits, project management, and CMS work. For a team of one, that workload is the whole problem.

The self-driving approach: automate the pipeline end to end

A self-driving content system runs the pipeline:

• Crawl your site and map existing coverage
• Find content gaps against SERPs and competitors
• Research topics and intent
• Write in your voice
• Publish on a schedule
• Learn from performance and adjust

That matches the multi-phase model described in mainstream AI marketing guidance, but removes the glue work that makes it fail in small teams (Salesforce, Airtable links above).

Keep only the hand-offs a solo marketer can sustain

The only sustainable hand-offs are:

• One-time calibration: positioning brief, voice pack, posting lanes
• Optional approvals: outline-only or final draft, with SLAs
• Monthly performance review: check the scorecard, pick next clusters, schedule retrofits

Everything else needs to be automated or templated.

Evaluate tools with four criteria (ignore feature lists)

Score tools on:

1. End-to-end publishing: can it publish into your CMS on schedule without copy-paste?
2. Voice consistency: does it hold tone and terminology across posts?
3. Analytics feedback loop: does it use Search Console and conversion data to change what it produces?
4. Permissions and approvals: can you run outline-only approvals safely?

If a tool fails any one of these, you will end up with another half-built system.

Where Highway fits

Highway is built for this exact constraint: consistent blog output without prompts, writer management, or CMS busywork.

It runs an autonomous pipeline: crawls your site, finds gaps, researches, writes in your voice, and publishes on a schedule. You calibrate once, then approve (or not) based on your risk tolerance.

The goal is not to become an AI power user. It is to publish weekly without hiring, without herding freelancers, and without living in Google Docs.

← Back to the blog