SPM
Successful Project Manager

Ship product. With AI in the room.

Agile software delivery and product management, augmented by Claude. Frameworks, templates, case-studies, blog, and resources — run by Mat Siems.

About Mat

One PM. One stack. Thirteen sites.

Mat Siems is an AI-augmented product manager and operator based in the UK. He runs a fleet of 13 live sites, a personal AI cockpit (AppAI), and a library of ~30 Claude Code skills — all as a single operator. SPM is the consulting practice that packages everything that works and delivers it to teams who want to move at the same speed without building the stack from scratch.

13
live sites under management
~30
Claude Code skills in the library
100%
AI-native delivery stack
1
human operator behind it all
The ultimate AI PM workflow

Zero to hero, idea to prod, red to green.

A TDD-shaped product loop where every gate is binary — tests pass or fail, UAT signs off or doesn't, security clears or blocks. No "kinda done" states, no waterfalls hiding inside sprints. SPM Agent runs the wheel; you review the diffs.

  1. 01
    Vision
    problem + outcome
  2. 02
    JSX
    clickable mockup
  3. 03
    PRD
    user outcomes
  4. 04
    Spec
    interfaces + AC
  5. 05
    Tests
    failing first
  6. 06
    QA fail
    red gate
  7. 07
    Build Vx
    agent writes diff
↓ red gate breaks the chain ↓ agent earns green ↓
  1. 08
    QA pass
    green gate
  2. 09
    UAT
    human review
  3. 10
    UAT updates
    triage P0/P1
  4. 11
    Security
    /security-review
  5. 12
    Prod
    canary → ramp
  6. 13
    Vx + 1
    loop close
The Ultimate AI PM Workflow — idea to prod, red to green, Vx → Vx+113 stages · TDD-shaped · every gate is binaryVisionproblem + outcomeJSXclickable mockupPRDuser outcomesSpecinterfaces + ACTestsfailing firstQA failred gateredBuild Vxagent writes codeQA passgreen gateUAThuman reviewUAT updatestriage P0/P1Security/security-reviewProdcanary → rampVx + 1 — telemetry + UAT triage feed the next visionLegendRed — tests fail, blocks buildGreen — tests pass, advanceTriage — UAT findings to next VxBrand — agent-owned stage010203040506070809101112
R

Red — fail first

Every feature starts as a failing test. If you can't write the failing test, you don't understand the feature yet — go back to the spec.

G

Green — earn the pass

Agent writes the smallest diff that flips the test green. Human reviews the PR. Refactor only after every test is green.

Vx → Vx+1 — close the loop

UAT triage + prod telemetry become next vision input. Most products die from forgetting this loop, not from failing it.

Solutions

Four products that ship the loop.

SPM Agent runs the project. AgentAI manages every agent in it. PersonAI gives them a face. AIOS is the cockpit you live in all day.

SPM Agent

Live

Manage any project with AI on autopilot.

A persistent project-manager agent that owns the loop from idea to prod — vision, JSX, PRD, spec, tests, build, QA, UAT, security, release — and ships the next version while you sleep.

  • Per-project skill scoped to your repo + your house style
  • Daily standup digest pushed to Slack/Teams
  • PBI grooming + acceptance-criteria authoring
Read more

AgentAI

Beta

AI agents management — the control plane for every agent you run.

A single cockpit to design, deploy, observe, and retire every Claude agent in your org. Skills, triggers, memory, cost, escalation policy — one pane of glass.

  • Agent registry with role + scope + owner per entry
  • Live invocation log with token + cost + latency per call
  • Skill marketplace: install, version, share inside the org
Read more

PersonAI

Beta

AI agent personas management — give every agent a face, a voice, and a memory.

The persona layer that sits under every agent — name, role, tone, knowledge cut, refusal posture, escalation chain. Cloneable, versionable, comparable.

  • Persona file: name, role, tone, refusal rules, escalation chain
  • Version history with diff + rollback
  • Clone-and-tune for new variants without forking the whole agent
Read more

AIOS

Live

AI Operating System for AI-preneurs — your whole stack, one cockpit.

The master workspace for solo operators and small teams running an AI-native business. Workspace UI, 13 Fox agents, Claude API proxy, templates, history, settings — all local-first, all yours.

  • V3 Master Fox workspace UI on localhost:3000/workspace
  • FastAPI sidecar on port 5400 for local-only operations
  • 13 Fox agents covering the AI-preneur surface area
Read more
Proof

Numbers from the loop, not slideware.

Everything below ships out of live engagements — anonymised but real. The point isn't the numbers; it's that they're typical of how the loop compounds when an operator runs it daily.

13

Live sites in the flexappdev fleet, all on the same skill stack.

30+

Reusable skills in the abc-* family — invoked daily.

POMs

25-min work units. Replace half-day meetings; outlive retainers.

0 → 6d

Median cycle time after a 90-day SPM engagement.

What you get

Frameworks, not vibes.

Concrete delivery playbooks, ADR templates, intake → proposal → shipped-product loops, and a skill library tuned to your stack.

How you buy

POMs, not retainers.

One POM = 25 minutes. Blocks from 5 to 20 POMs, plus Managed Agents from £1 / agent-hour. No minimums, no lock-in.

Who runs it

One operator. One stack.

Mat Siems. 13 live sites, the AppAI cockpit, the ABC skill family, and a fleet of agents — all the same stack you'll be using.

Common questions

Things people ask before booking.

Is this consulting, training, or a product?

All three, packaged as POMs. A POM block buys you Mat's attention and the skill stack. You can spend it on hands-on delivery, framework coaching, or product setup — same rate, you decide the mix per call.

How fast can we start?

Booked free POM today, kickoff inside a working week. Most engagements ship their first PBI before week two — that's the point of the loop. If you need a 6-month statement-of-work negotiation, this isn't for you.

Do I need to be technical?

No. Half the engagements are with non-technical CPOs and founders. Mat operates the stack; you operate the product. The agents handle the parts that used to require an engineer in the room.

What if our team uses GPT / Gemini / our own LLM?

The skill format is provider-neutral. We default to Claude (Opus 4.7 / Sonnet 4.6) for cost and quality, but the same agents run against other models if your stack requires it.

Is there a minimum commitment?

No retainer, no contract. Smallest unit is a 5-POM block (£500). You scale up if it's working, walk if it isn't. Most clients land at 20 POMs/month within 60 days — but that's their call, not ours.

What happens to the IP?

Yours. Every skill, agent, PBI, and ADR we co-write belongs to you the moment it lands in your repo. No license. No clawback. The only thing Mat keeps is the abstract pattern — and even that's on GitHub.