Two months of foundation work. A checkpoint at day 30 where you can stop, and a full decision at day 60.
A full marketing operation runs every channel at once. We are a small team, so we are not promising that.
Grounded in the three research reports delivered 2 to 11 August 2026: competitive analysis, market research, and competitor content framing.
The whole plan on one slide. Every part after this goes deeper on one piece of it.
Practitioners find Marrow on the channels they already use, install it in minutes, get proof they can forward to their lead, and name us when leadership picks a vendor. A smaller capture slice runs alongside it, so buyers already shopping can find us without waiting for a practitioner.
Proof. Policy is a rule. Proof is evidence. No competitor in the category uses the word, so every piece of content ladders back to it.
Two audiences, two funnels, one handover between them. Production starts in week 2.
All Phase 1 content points here. They can decide alone, with no procurement.
| Titles | AI Platform Engineer · AI Infrastructure Engineer · AI Automation Engineer · AI Agent or Applied AI Engineer · MLOps and Platform Engineer · Internal tooling or DevEx lead · Solutions architect at an AI consultancy · Technical founder running an agent fleet |
| Where they are | X · GitHub · Discord, in agent, MCP and Claude Code communities · YouTube · docs and changelogs · inside the harnesses themselves, like Claude Code, Codex, Cursor and MCP clients |
| What they need | "Ship agents that do not quietly fail, and be able to show my lead this is safe to run unattended." |
| How they buy | Try first, read docs, check GitHub, decide alone. The cost of trying has to be near zero. |
| What moves them | A real failure they recognise, a fix they can run today, and proof the tool is not hand waving |
The destination, not the starting point. In Phase 1 we build the material that meets them. We do not chase them.
| Titles | CTO · VP Engineering · COO · Head of Risk or Compliance Officer · Head of Automation or Digital Transformation · Director of Operations · clinic practice owner or managing partner |
| Where they are | LinkedIn · email · vendor shortlists and internal evaluations · peer referral · industry press · increasingly, whatever an AI assistant tells them |
| What they need | "Adopt AI agents without creating an unauditable liability, and pick a vendor I can defend to my board." |
| How they buy | Shortlist, evaluate, involve risk and procurement. Weeks to months. They rarely start the search themselves. |
| What moves them | Someone internal they trust saying "we should use this", plus material that survives a compliance read |
The buyer's top of funnel is mostly an output of the practitioner's bottom of funnel. The only two entry points that do not depend on it are AI search and the comparison page. Part 2 explains why we fund those two separately. They are the only way in that does not need a practitioner to find us first.
Most content plans skip this step. It decides whether any of the work reaches a buyer at all.
Every practitioner asset ends in something they can paste into a thread with their lead. A proof output, a readiness result, a before and after.
When they go looking for how to justify it internally, the lead magnets and the readiness assessment are already sitting there, behind an email form.
The download tells us their industry and their intent. That is the handoff to Albert or Jenny. A warm lead who raised their hand, not a cold list.
TechiesLab owns everything up to the handoff. We do not do the outreach. We make sure the person on the other end already asked for it.
Albert wants Marrow in front of CTOs, COOs and Heads of Risk. That is the right destination. It is not a reachable starting point today.
Outbound reps, a territory, a follow up cadence. We have none of this.
Credo AI runs its whole motion on Forrester and Gartner partnerships. That access costs money we do not have.
Customer logos, vendor list placements, funding announcements. Competitors have 64 to 189 million raised. Marrow has documentation.
A top down plan would mean polished LinkedIn posts aimed at people with no reason to open them. The one door open to us is the practitioner already running agents inside those companies.
Four competitors, four lessons. Each one either confirms a choice in this proposal or rules something out.
| Competitor | What they do | What it means for us |
|---|---|---|
| Credo AI | Board level framing, regulatory authority, zero developer content by choice. See their Forrester co-hosted webinar. | Top down works, but it runs on analyst access and budget. We cannot copy this. |
| Galileo now Cisco | Very active on X with technical demos. Blogs that open with a real production incident, like their governance framework post. Grew developer first. | Developer first content works in this exact market. This is the closest template to what we are proposing. |
| Kosmoy and Straiker | Both farm evaluation search with comparison posts: Best AI Agent Governance Platforms 2026 and Top 7 AI Agent Visibility and Governance Platforms. | Buyers are already searching. There is demand to capture today, not only demand to create. |
| Taktile | One vertical only. Use case deep dives, customer results, G2 badges. | Vertical material converts. We can copy the structure, but we have no customer results, so we use scenarios instead. |
Credo, Zenity, Straiker and Kosmoy all explicitly avoid developer content. That gap is the opening. The research also found a messaging gap, which we use in Part 3.
These are not competing strategies. They do different jobs, cost different amounts, and pay back on different timelines.
| Capture demand that exists | Generate demand that does not | |
|---|---|---|
| The reality | People already search "best AI agent governance platform". Kosmoy and Straiker farm that traffic today. | Nobody searches for "proof layer" yet. The category has no vocabulary. |
| The job | Get Marrow onto those lists, and into the answers AI assistants give. | Make practitioners feel the problem, then own the word that names it. |
| Where | Comparison and alternatives pages, GitHub, MCP listings, docs depth, third party roundups, llms.txt and .well-known files | X demos, incident to proof stories, ecosystem demos, industry scenarios |
| Trade off | Fast, cheap, high intent. Small volume, and easy for a funded competitor to outspend. | Slow, compounding. This is the part nobody can copy quickly. It is the moat. |
| Phase 1 split | About 20% | About 80% |
They meet in the same funnel. Generation creates the practitioner who installs Marrow. Capture catches the buyer who is already shopping. The practitioner is the reason Marrow survives that buyer's shortlist.
Google Cloud gives developers free credits because the developer is the person asked "what should we use" when the company picks a stack. Datadog and Sentry grew the same way. The research names this as a live opportunity for Marrow.
Through content on the channels they already use.
One npx command on their own agent setup. Minutes, not a procurement cycle.
Output they can forward upward to their lead.
When the CTO or Head of Risk asks who to use, their own engineer already has the answer.
The Phase 1 goal: get Marrow's name into the AI agent governance consideration set, carried there by people who have actually used it.
| Channel | Status | Demand type | What it does |
|---|---|---|---|
| X | Primary | Generate | Discovery. Product demos, ecosystem demos, incident to proof stories. |
| GitHub | On | Both | Where practitioners check we are real, and where evaluators land from search. |
| Discord | Light | Generate | Hold the people who tried it. Capture email on join. |
| Website | On | Both | Lead magnets behind an email form, plus the comparison and alternatives pages. |
| Listings and AI search | On | Capture | Glama MCP listing, docs depth, llms.txt and .well-known files. Where AI assistants read us from. |
| On | Both | Segment by industry and intent. The only list we own. | |
| Light | Capture | Trust layer. Makes Albert's manual 1:1 outreach land. | |
| YouTube | Mirror | Generate | Re-host the X demos so they are findable. Not grown as its own channel this phase. |
| Reddit, webinars, workshops | Parked | None | Not ready. Revisit at a later gate. |
Marrow currently has no X, LinkedIn or YouTube presence at all. Setting these up is part of Phase 1.
The research checked every competitor's messaging. The word appears in none of them. Galileo comes closest with "immutable decision rationales" and still avoids it. Everything on the next three slides is built on this line.
Policy says "do not do X."
Proof says "here is the evidence X did not happen, under these controls, with this outcome."
Policy is a rule. Proof is evidence. Auditors need evidence.
The window is 12 to 18 months before a funded competitor adds "proof" to their feature list and takes the word. That is the argument for starting the content now rather than next quarter.
| For practitioners X · GitHub · Discord · YouTube | For buyers LinkedIn · email · website · lead magnets | |
|---|---|---|
| Voice | Personal, first person, hands on. Closer to consumer than enterprise. | Operator to operator. Measured and specific. |
| Format | Short demo clips, screen recordings, before and after, code | Guides, scenario walkthroughs, checklists, readiness material |
| Hook | A real incident, then the proof that would have caught it | "When your auditor asks you to prove the agent followed policy, what do you show them?" |
| Never | Industry playbook framing. "The playbook for clinics" means nothing to someone who has not felt the problem yet. | Hype tone, meme formats, or any number we cannot source. |
| Call to action | Try it on your own setup | Download the material, email required |
The incident to proof hook is taken straight from the whitespace list. Galileo does incident to eval, Lakera does incident to defense. Nobody does incident to proof.
One formula: a specific agent action reported success, something quietly failed, here is the gate that caught it. Personal, first person, no industry framing.
| Material | Hook variations |
|---|---|
| Incident to proof clip X, YouTube mirror | "My deploy agent said the release shipped. It shipped to staging. Here is the permit that would have stopped it." "Six agents ran overnight. Five reported done. I can prove three." "A tool call returning 200 is not a completed obligation. Here is what closing an outcome actually looks like." |
| Ecosystem demo X, YouTube mirror | "Marrow inside Buzz: what governance looks like when it is not a separate dashboard." "We put a pre-action gate inside an agent chat app. Here is what the team actually sees." |
| Feature launch X, changelog | "New: signed action permits. The 40 second version." "Break glass now needs a reason and expires in five minutes. Here is why that matters at 3am." |
| Install prompt GitHub, Discord, docs | "One npx command. Then watch it block the thing you did not want it to do." "Your agents lie to you politely. This is the receipt." |
Same idea, different register: risk and defensibility. Quotable in a meeting, survives a compliance read, no number we cannot source.
| Material | Heading variations |
|---|---|
| Lead magnet Document, research or freebie | "AI Agent Setup for Clinics: controls, evidence, and what to check before you turn it on" "Agent Governance Readiness: the seven baseline controls" |
| LinkedIn post | "When your auditor asks you to prove your AI agent followed policy, what do you show them?" "Policy is a rule. Proof is evidence. Auditors need evidence." "Your model risk framework covers models. What covers the agents acting on model output?" |
| Email subject | "The question your auditor has not asked yet" "Clinic agent controls, the short version" |
| Checklist freebie Website, email | "The Pre-Action Gate Checklist" "What to record before an agent touches a patient record" |
These are written against queries people already type, and against what AI assistants read when someone asks them to recommend a vendor.
| What they are looking for | Page heading |
|---|---|
| Category evaluation | "Best AI agent governance platforms in 2026, compared" |
| Head to head | "Marrow vs Credo AI" · "Marrow vs Kosmoy" · "Marrow vs Zenity" |
| Displacement | "Independent alternatives, for teams whose agent governance vendor just got acquired" |
| How to, per runtime | "How to govern Claude Code" · "How to govern Codex" · "How to govern Cursor" |
| Protocol first mover | "How to govern MCP servers" |
| Industry entry | "AI agent controls for clinics" · "for insurance claims" · "for finance operations" |
One condition on the comparison pages: they have to be genuinely fair. Kosmoy's and Straiker's versions read as self serving, and a company selling proof gets held to a higher standard than one selling a gateway. Accuracy here is a positioning asset, not just good manners.
Four lines we commit to. One that depends on Albert.
A feature ships, a short demo follows. Templated and repeatable. The backbone of the X cadence.
Marrow inside Buzz, inside Hermes, inside whichever agent native product comes next. Borrows their audience.
Built from the existing "Marrow lives inside Buzz" scenarios, recut for one industry. Feeds the lead magnet and the email list.
A fair comparison page, an alternatives page, docs and MCP listing optimisation. Small effort, catches buyers already evaluating.
The strongest trust builder we have, and the one we cannot guarantee. It needs Albert to change how he documents his work. Upside, not a deliverable.
One worked example, so the writing rule on the last slide is concrete.
| Step | Where | What happens |
|---|---|---|
| Hook | X | A real scenario, told personally. "My scheduling agent reported 12 confirmations sent. Three never went out. Here is the gate that caught it." No industry framing, no playbook language. Just a failure a practitioner recognises. |
| Try | GitHub, docs | They install it on their own setup and watch the gate work. |
| Search | Website, AI search | Now interested, they look for their own context. "We do this for clinics. Is there something for us?" |
| Convert | Website | They find AI Agent Setup for Clinics. Button: view the full material. Email required. |
| Segment | That address enters the list tagged industry: clinics. We now know who they are and what they want. | |
| Nurture | Clinic specific sends. Later, a clinic specific session if the segment grows. | |
| Handoff | Marrow | Albert, Jenny or a future rep reaches out. TechiesLab's work ends here. |
Week 1 is setup. Content goes live in week 2.
Two exits, not one. At day 30, 14 September, you check the machine is running as promised and can stop there, having spent one month. At day 60, 14 October, we look at real evidence together and decide about Phase 2. You are never buying more than a month ahead.
Inbound lead generation is ours, on both sides. Practitioners through Discord and installs, buyers through gated industry material, up to the tagged lead. The conversation after that belongs to Marrow.
Set at what TechiesLab can hold without dropping quality. These are floors, not targets. We would rather beat them quietly than explain a miss.
| Rhythm | Committed output | Notes |
|---|---|---|
| Daily working days | 1 monitoring pass across X, Discord and GitHub | Anything needing Albert is escalated the same day. No daily posting commitment. |
| Weekly from week 2 | 3 X posts · 2 LinkedIn posts · 1 production asset shipped · 1 Friday progress note | The production asset alternates between a demo clip and a written piece. |
| Monthly | 1 planning session · 1 written report against the baseline · 1 major asset | Month 1 major asset is the first lead magnet drafted. Month 2 is that magnet shipped plus the capture pages. |
| Month 1 15 Aug to 14 Sep | 9 X posts · 6 LinkedIn posts · 2 demo videos · 1 email send | Three weeks of cadence, since week 1 is setup. |
| Month 2 15 Sep to 14 Oct | 12 X posts · 8 LinkedIn posts · 2 demo videos · 3 email sends | Plus the full lead magnet set and both capture pages. |
| Phase 1 total | 21 X posts · 14 LinkedIn posts · 4 demo videos · 4 email sends · 1 lead magnet set · 2 capture pages | Plus 8 weekly notes and 2 monthly reports. |
Not committed: daily posting, same day turnaround on ad hoc requests, or weekend coverage.
Phase 1 is not expected to produce pipeline. It is expected to produce a working system and a first honest read on whether the message lands. Anyone promising revenue in month one of a category with no vocabulary yet is guessing.
What we need from Albert: X API access so posting can be automated · about an hour a week for demo input and fact checks · the industry decision · and, when possible, 2 to 3 design partners signed with permission to publish their own results later. The research names zero customer proof as Marrow's biggest weakness, and that permission is much harder to add after signing.
| # | Step | Owner | When |
|---|---|---|---|
| 1 | Approve scope and pick one industry to lead with: clinics, finance or insurance | Albert | Fri 15 Aug |
| 2 | Kickoff and access handoff: X, GitHub, Discord, LinkedIn, website, email tool, analytics | Both | Mon 18 Aug |
| 3 | X API access arranged for posting automation | Albert | Week 1 |
| 4 | Channel accounts set up and branded | TechiesLab | Week 1 |
| 5 | Baseline snapshot plus audience map and channel matrix delivered | TechiesLab | Week 1 |
| 6 | Content production starts. First posts and first demo live. | TechiesLab | Week 2, from 22 Aug |
| 7 | Email capture and industry segmentation live | TechiesLab | Week 3 |
| 8 | Day 30 checkpoint. Running as promised? Stop here, or continue. | Both | Mon 14 Sep |
| 9 | Lead magnet set shipped: material, landing page, form, sequence | TechiesLab | Week 5 to 6 |
| 10 | Comparison and alternatives pages published, listings optimised | TechiesLab | Week 7 |
| 11 | Day 60 gate. Decide together about Phase 2. | Both | Wed 14 Oct |
Everything in Part 1 comes from these. All links open in a new tab.
| Competitor | Position | Materials |
|---|---|---|
| Galileo (Cisco) | Developer first. The closest template to this plan. | X · YouTube · Governance blog · Events |
| Credo AI | Governance authority, analyst led, top down only. | Forrester webinar · Webinars hub |
| Kosmoy | Best funded pure play. Comparison SEO engine. | Best Agent Governance Platforms 2026 · Pricing |
| Straiker | Security challenger. Also farms comparison search. | Top 7 Platforms · Blog |
| Zenity | Agent security, CISO targeted, category creation play. | Governance checklist for CISOs · X |
| Lakera (Check Point) | Threat intelligence. Incident to defense framing. | Blog · X |
| Taktile | Vertical proof. Fintech only, customer results led. | Blog · X |
Marrow's own surfaces: getmarrow.ai · Docs · GitHub · Glama MCP listing · Governance Readiness Assessment
Month 1 builds it, month 2 runs it, and you can stop at day 30 if it is not being built. Capture the demand that already exists, and start building the demand nobody has claimed. The research says the window on that second one is 12 to 18 months.