New Service AI Engineering

Install marketing AI inside your company.

We take the 20+ agent infrastructure we built to run Hedgehog's own marketing — content, ads, SEO, reporting, the whole machine — and install a custom version of it inside your company's systems. For in-house marketing teams who’d rather own their AI than rent it.

From $10,000 + GST

One-off build. Then from $500 + GST / month for ongoing engineering, updates, and feature requests.

your-business — agents/
# Your marketing stack, running on your infra
$ ls agents/
ceo-orchestrator google-ads-mgr
meta-ads-mgr content-agent seo-agent
reporting brand-qa +13 more
$ ceo run morning-ops
Shipped 3 Google Ads optimisations
Queued 2 Meta creative tests
Drafted weekly report — awaiting QA
Slacked lead: “1 thing needs your eyes”
Why this exists

Every SaaS is selling you AI. Almost none will install it in your company.

Most “AI marketing” offers today look the same: vendor builds something on their infrastructure, charges you a monthly subscription, and you rent access. When you cancel, the AI leaves. Your data trains their product. You never own the thing.

We built it the other way. Over the last 12 months we’ve shipped a working multi-agent marketing operation that runs inside Hedgehog — content, paid ads, SEO, reporting, brand QA, the lot. Now we can install that same system, customised to your brand and your marketing ops, inside your company.

1

You own it

Installed on your infrastructure. Your repos, your secrets, your data. When we’re done, the AI is yours. Full stop.

2

Custom to your team

Every agent is tuned to your brand voice, your products, your customers, your reporting needs. Not a generic template with your logo on it.

3

Real infrastructure

Built on the same stack that runs ours. Claude, Anthropic SDK, Google Workspace, Slack, your ad platforms, your CRM. Production-grade, not a demo.

The Agent Catalog

20+ agents built for in-house marketing teams. Picked, tuned, and installed based on how your team works.

Not every company needs all of them. The scoping call is where we figure out which ones actually matter for your team. Here’s the menu:

CEO Orchestrator Receives every brief, routes work, runs peer review between agents.
Strategist Writes briefs before execution. Checks output against the brief.
Meeting → Tasks Reads Gemini / Fireflies / Otter transcripts. Turns action items into tasks in your ClickUp / Asana / Notion.
Planning Agent Drafts monthly marketing plans, content calendars, campaign timelines.
Content Agent Drafts blog posts, landing pages, product copy — on-brand.
Email Agent Nurture sequences, broadcasts, lifecycle flows, abandoned cart.
SEO Agent Keyword research, metas, topical authority plans, on-page.
Ad Copy Agent Writes Google, Meta, LinkedIn ad variants tuned to your campaign angle.
Google Ads Autopilot Daily bid adjustments, keyword pruning, budget pacing, ad copy rotation — on guardrails you set.
Meta Ads Autopilot Auto-creative testing, audience iteration, performance-based scaling on your Meta accounts.
Creative Director Briefs visual work, grades designs against the brief, calls the shots.
Graphic Designer Generates concept imagery, ad creative, social graphics via image gen + Canva.
Brand QA Checks every deliverable against your brand guidelines before it ships.
Copywriter QA Strips AI slop. No flat, generic, corporate tone — ever.
Data QA Cross-checks every number in a report against GA4, GSC, Ads, CRM.
Reporting Agent Pulls GA4, GSC, ad platforms, CRM. Drafts weekly / monthly reports.
Anomaly Agent Flags spend spikes, conversion drops, ranking drops, traffic crashes.
Competitor Intel Tracks competitor ads, organic pages, pricing, and flags what’s changed.
Social Listening Watches for brand mentions, trending topics, and competitor news.
Tracking QA Monitors GA4, GTM, pixels, conversions. Flags broken tracking immediately.
Ops Engineer Builds new internal tools and integrations on request. Ships fixes.

Want one we don’t have? That’s what your maintenance tier covers — feature requests, new agents, integrations with whatever tool your team adopts next.

Signature capability

Ads Autopilot — your media spend, managed overnight.

The two agents we get asked about most. Most clients use this as their actual first reason to install the stack — because ad spend is where waste shows up fastest and where agents pay for themselves quickest.

Google Ads Autopilot

  • Pauses zero-converter keywords before they burn another day of budget
  • Reallocates spend into winning ad groups based on rolling 7/14/30-day performance
  • Rotates RSA headlines & descriptions when any asset dips below a threshold
  • Watches Quality Score drops and surfaces the exact keyword/ad pairs to fix
  • Builds negative keyword lists from the search terms report every morning
  • Flags Performance Max and Shopping anomalies before they compound

Meta Ads Autopilot

  • Runs creative testing cycles — kills losers, scales winners, briefs new variants
  • Iterates audiences: lookalikes, interests, custom audiences — tested and pruned weekly
  • Monitors creative fatigue via frequency & CTR decay, queues refresh before results crash
  • Balances budget across adsets on CPA / ROAS — whichever your guardrails prioritise
  • Builds static + video variant briefs for the Graphic Designer agent to execute
  • Flags Meta policy issues & account-health red flags the second they surface
🛡️
Always on a guardrail. You set the ceilings — max daily spend, max bid increases, approved account access. The autopilot runs inside those bounds and asks you for approval the moment it hits one. No surprise bills, no silent changes.
The pipeline, not the demo

Every task routes through a CEO agent. Every output gets peer-reviewed before it reaches you.

This is how a real multi-agent system works — not one model doing everything, but a hierarchy where a CEO orchestrator receives briefs, routes work to specialists, and makes them check each other’s output before anything hands back to a human. Here are three typical workflows:

The CEO is the front door. Everything starts with her.

You don’t talk to 20 agents directly. You brief the CEO — in Slack, email, or a task — and she decides: who owns this, what context they need, who should review the output, and when it’s ready to come back to you. It’s the difference between an AI tool and an AI team.

The Prep or Do It rule

The CEO decides: can the agents ship this, or do they hand you a prep kit?

Every task that lands on the CEO gets sorted into one of two buckets. This is the single most important rule in our stack — it’s what stops agents from over-reaching, and it’s what makes you trust the system.

✅ Do It

The agents have the skills → they ship it.

Writing a blog post, drafting ad copy, optimising bids, building a weekly report, flagging anomalies. Anything with clear guardrails and agents trained for the job. The CEO routes, they execute, QA reviews, you approve.

Example: “Ship a blog about our new pricing” → Content Agent drafts, QA reviews, CEO delivers a ready-to-publish draft.

📝 Prep It

New territory → the agents prep you for it.

Something the agents can’t (or shouldn’t) execute alone — board decks, new hire briefs, strategic pivots, customer escalations. Instead of failing or faking it, the agents build you a prep kit: context, options, draft materials, risks.

Example: “We need a customer retention strategy” → agents pull churn data, competitor research, and draft options — you walk into the meeting prepped, not starting from zero.

Example 01 — Weekly content

“Ship a blog post on our new feature by Friday.”

Brief lands in Slack → routes to the CEO orchestrator.

  • CEO reads the brief, pulls your brand guidelines + product context, assigns to Strategist.~30s
  • Strategist writes a detailed content brief — angle, outline, SEO target, CTA. Hands to Content Agent.~2 min
  • Content Agent drafts the post. Submits to QA.~5 min
  • Copywriter QA reviews for AI slop, flat writing, lazy structure. Kicks back 3 edits.~1 min
  • Brand QA checks against brand voice, tone, disclaimers. Passes.~1 min
  • CEO packages the final draft + SEO metas and drops it in your inbox for approval.~30s
Example 02 — Automated ads

“Keep our Google & Meta ads optimised.”

Runs on a schedule. No brief needed.

  • CEO triggers daily ads sweep.Every morning
  • Google Ads Manager pulls yesterday’s performance. Pauses 2 underperforming keywords, shifts 8% budget to the best-converting ad group.~90s
  • Meta Ads Manager checks creative fatigue. Queues 3 new variants for the winning adset.~2 min
  • Data QA validates the changes against your guardrails (no over-bidding, no audience collapse). Approves.~30s
  • Anomaly Agent spots a conversion tracking gap on one campaign. Flags to CEO.~1 min
  • CEO posts a one-line summary in your Slack: “3 optimisations shipped. 1 tracking issue flagged — see thread.”~30s
Example 03 — Monthly report

“Get the October performance report ready for the exec team.”

Runs on the 1st of the month — or on demand.

  • CEO kicks off the monthly pipeline at 7am.1st of month
  • Reporting Agent pulls GA4, GSC, Google Ads, Meta Ads, your CRM. Drafts narrative sections.~4 min
  • Data QA cross-checks every number against source-of-truth dashboards. Flags one GA4 discrepancy.~2 min
  • Reporting Agent fixes the discrepancy, adds a note on the anomaly.~1 min
  • Brand QA checks formatting + exec-ready tone. Passes.~1 min
  • CEO drops the polished .docx in your Drive and pings your marketing lead.~30s
Example 04 — Campaign kickoff

“Launch a new product campaign in 2 weeks.”

Big brief. Routes through the full stack.

  • CEO receives the brief, breaks it into workstreams, assigns owners.Day 1
  • Strategist writes the campaign brief. Planning Agent builds the timeline.Day 1
  • Competitor Intel scans the space. SEO Agent picks target keywords. Creative Director writes the visual brief.Day 2–3
  • Content Agent drafts landing page + blog. Ad Copy Agent writes 12 ad variants. Graphic Designer concepts 6 creatives.Day 4–6
  • Brand QA + Copywriter QA cycle. Feedback round-trips until it ships-ready.Day 7–8
  • CEO packages the full campaign kit for your approval and briefs the ads agents for launch day.Day 9
Installed vs Hosted

Why own it instead of renting it.

Most "AI for marketing" subscriptions are just an API wrapper + someone else’s prompts. You pay monthly forever. You don’t learn anything. You don’t keep anything.

What you get Hosted AI tool AI Engineering (installed)
Where it lives Their servers Your infrastructure
Your data Trains their product Stays with you
Customisation Whatever their roadmap allows Anything you want, built for you
If you leave Lose access, start again Keep everything, keep running
Cost over 3 years $200–$2,000/mo × 36 = $7k–$72k, and you own nothing $10k build + $6k–$36k maintenance = $16k–$46k, and you own everything
Feature requests "We’ll add it to the backlog" Shipped in your next monthly update
The Build Process

From scoping call to shipped infrastructure in ~6 weeks.

Step 01

Scoping call

We map your current marketing ops, workflows, tools, pain points. Then we pick the 5–10 agents that’ll move the needle first.

Week 1
Step 02

Infrastructure setup

We provision your environment, wire up your Anthropic + Google + CRM + ads accounts, and set up the agent runtime in your stack.

Week 2
Step 03

Agent customisation

We take each agent from our stack and retune it for your brand, your voice, your data, your reporting needs. No generic prompts.

Week 3–5
Step 04

Training + handover

We train your team on how to brief agents, review their work, extend the system. You leave week 6 with a running operation.

Week 6
Investment

Transparent pricing. Every project scoped individually.

Build — one-off
From $10,000 + GST
Scoped based on number of agents and integration complexity. Typical range: $10k–$25k. Larger installs ($25k–$50k) quoted on scope.
  • Scoping call + system design
  • Full infrastructure setup in your stack
  • 5–10 custom-tuned agents installed
  • Integration with your CRM, analytics, ad platforms
  • Team training + documentation
  • 90-day warranty on the initial build
Maintenance — pick a tier, change anytime
Light
Care & Feeding
$500
+ GST / month

Monthly health check, updates when we ship improvements to core agents, small fixes. No new builds.

Heavy
Full Partner
$3,000
+ GST / month

Everything in Medium, plus ongoing architecture work, 3–5 features per month, and a dedicated engineer on standby.

Tiers are a floor, not a ceiling. Scale up for a busy feature quarter, scale down when things are steady. Cancel any tier with 30 days’ notice.

Good fit / Not a fit

This is not for everyone. And that’s the point.

You’re a good fit if…

  • You run marketing in-house for a real company — B2B SaaS, ecommerce, services, manufacturing, whatever.
  • You’ve got a marketing team of 2–20 people, or a single marketing lead who’s buried in execution.
  • You’re already spending serious money on ads, content, and tools — and seeing the inefficiency you want to fix.
  • You want AI installed permanently inside your company, not another SaaS subscription that rents it to you.
  • You have — or can get — a technical point-of-contact (internal dev, fractional CTO, or contractor) to work with us during the build.
  • You’re ready to invest in a real system, not a demo.

You’re not a fit if…

  • You’re a marketing agency looking to install this for your clients — that’s a different scope (talk to us separately).
  • You’re looking for a cheap AI chatbot you can bolt on in an afternoon.
  • You don’t have ongoing marketing work for agents to plug into.
  • You want us to run your marketing for you — that’s what Hedgehog in Your Corner is for.
  • You can’t commit any internal capacity to the 6-week build phase.
  • You want to “try” AI before committing. This is a real investment, not a sandbox.
FAQ

Questions we get a lot.

What does "installed in my business" actually mean?

We deploy the agent infrastructure inside your own accounts — your Anthropic API key, your Google Cloud project, your GitHub repo, your Slack workspace. The agents run on your infrastructure, not ours. You own the code, the configuration, and the data. We’re the builders; you’re the owner.

Do I need to be technical?

You need a technical point-of-contact on your side — someone who can give us access to your accounts, review pull requests, and do basic ops work once we hand over. That could be an internal developer, a contractor, or someone on your team with engineering chops. It doesn’t have to be the business owner.

What if I want more than 10 agents?

The $10k floor buys you 5–10 agents well-tuned for your core ops. Bigger installs with 15+ agents, complex integrations, or custom workflows typically scope at $25k–$50k. We’ll scope on the call — no hidden charges.

What do the maintenance tiers actually cover?

Light ($500/mo) is care and feeding — monthly health checks, updates when we improve our core agents, small bug fixes. No new builds. Medium ($1,500/mo) is the sweet spot for most clients: everything in Light, plus ~1–2 new agents or significant feature builds per month. Heavy ($3,000/mo) is for clients with a busy feature pipeline — ongoing architecture work, 3–5 features per month, a dedicated engineer. You can move between tiers with 30 days’ notice.

Can I cancel maintenance and keep the agents?

Yes. Everything we install is yours. If you cancel maintenance, your agents keep running — you just won’t get our ongoing builds, updates, or support. Most clients stay because the feature pipeline is worth it, but you’re never locked in.

How does this compare to hiring an in-house developer?

An in-house mid-level developer costs ~$120k/year fully-loaded. We cost $10k upfront + $6k–$36k/year depending on tier. We’re not a replacement for a dev team — we’re a specialist crew for one specific job: installing marketing AI. Once the install is done, a less-senior internal person can maintain it with our support.

What tech stack do you use?

The Anthropic Claude SDK for the model layer. Node.js / Python / TypeScript for runtime. Google Cloud or AWS for infrastructure. ClickUp + Slack + Google Workspace for orchestration. We adapt to your stack where it makes sense, but the core is proven tech that will still be working in 5 years.

Can I see a demo of your agent stack before I commit?

Yes. The scoping call is where we walk through what we’ve built for Hedgehog, show how the agents interact, and talk about which ones would map to your business. It’s free.

Want to see what this could look like for your business?

Book a free 45-minute scoping call with Josh. We’ll walk through our stack, map it to your ops, and give you an honest answer on whether it’s worth building.

← Back to homepage