Question 1

What is an AI agent and how is it different from a chatbot?

Accepted Answer

An AI agent is software that pursues goals autonomously — planning, using tools, and acting without constant human direction — whereas a chatbot only responds to prompts. In The Agentic Enterprise, Morgan Blake frames the difference as three superpowers chatbots lack: goal decomposition, tool use, and persistent autonomous operation. Blake devotes a full section to 'The Three Superpowers (And Why Your Chatbot Doesn't Have Them).' A chatbot waits for you to type; an agent receives a goal, breaks it into steps, executes those steps with real tools (browsers, APIs, file systems), and reports back. Ethan Mollick's framing, quoted in the book, is that the critical shift is not intelligence but agency — 'the capacity to take actions, use tools, and pursue goals without constant human direction.' The book's protagonist Dave discovers this in a fourteen-second demo where three agents process an insurance form while he watches. The practical consequence: chatbots answer questions, agents complete workstreams. That distinction underpins every system the book builds, from the CEO Agent that decomposes strategy to the Finance Agent that reconciles invoices at 3 AM without being asked.

Question 2

What is an agentic enterprise?

Accepted Answer

An agentic enterprise is a business where autonomous AI agents handle execution under human strategic direction — even a one-person business run from a spare bedroom. Morgan Blake's The Agentic Enterprise defines it as a human orchestrator plus a coordinated team of specialist agents operating under a four-layer governance model. Blake's definition rests on Sandeep Saini's Agentic Operating Model: four layers (Cognitive, Coordination, Control, Governance) that let autonomous systems scale safely. The book's central claim is that this architecture 'scales identically from laptop to enterprise' — LinkedIn's multi-agent support system and Dave's five-agent spare-bedroom operation share the same pattern: one coordinator, multiple specialists, human oversight for edge cases. Dave's enterprise comprises five agents: a CEO Agent for strategy, an HR Agent for provisioning, a Finance Agent for bookkeeping, a Marketing Agent for pipeline, and an IT/Security Agent enforcing zero-trust. The human supplies what the book calls the 30%: judgment, domain knowledge, and client relationships. The result is a business with enterprise-grade functions at roughly £4,700 per year in compute instead of £200,000 in salaries.

Question 3

How can AI agents be used in a small business?

Accepted Answer

AI agents can run a small business's strategy, operations, finance, marketing, and security as five coordinated specialists. In The Agentic Enterprise, Morgan Blake shows a solo operator deploying exactly that team — CEO, HR, Finance, Marketing, and IT/Security agents — on free open-source tools from a spare bedroom. The book maps each business function to an agent with a concrete open-source toolkit. The CEO Agent (CrewAI's Manager pattern) decomposes goals and delegates. The HR Agent (Ansible, Docker Compose, Portainer) provisions new agents in forty-seven seconds. The Finance Agent (Firefly III, GnuCash, FinGPT) categorises transactions, reconciles invoices, and processes forms in nightly batches — 340 forms between 2:47 and 4:03 AM at 3.8 seconds each. The Marketing Agent (Mautic, Plausible, Chatwoot) runs nurture sequences and A/B tests. The IT/Security Agent (Wazuh, Trivy, Falco, OpenBao) enforces zero-trust. Blake's revenue models for monetising this: Agent-as-a-Service, Productised Automation, Consulting with Agent Delivery, and Content/Research Agencies. Dave reaches £2,487 monthly agentic revenue against a £2,400 salary.

Question 4

Will AI agents replace human workers?

Accepted Answer

AI agents replace execution tasks, not human judgment — the book's 70/30 Rule holds that AI handles roughly 70% of work while the human's 30% (judgment, relationships, ethics) becomes more valuable, not less. Morgan Blake's The Agentic Enterprise argues agents augment the orchestrator rather than replace them. Blake is direct: 'You are not replacing humans. You are augmenting yourself — replacing the parts that were being wasted on Sunday VAT returns and eleven PM email spirals.' The 70/30 Rule, drawn from Dr. Laura Weis's research, structures every chapter: agents execute, humans decide. Stanford and CMU research quoted in the book found agents deliver a 96% cost reduction but are 'fundamentally different from humans,' not cheaper substitutes — they optimise within parameters but cannot negotiate between them. When Dave's Marketing Agent proposes £50 per batch and his Finance Agent demands £85, only Dave can resolve the conflict (£65, split the difference). The book's M-shaped professional model positions the human as the integration axis connecting multiple expertise domains — the role no agent can fill.

Question 5

How much does it cost to run AI agents for a business?

Accepted Answer

Running a five-agent AI workforce costs roughly £4,700 per year in compute — about £392 monthly — versus approximately £200,000 for equivalent human roles, a 97.6% cost reduction. The Agentic Enterprise by Morgan Blake builds the entire stack on free open-source tools and a consumer laptop. Blake itemises the economics in Chapter 13: five agents at roughly £940 per agent per year covers compute, inference, and cloud hosting if you scale beyond the laptop. The software bill is £0 — CrewAI, AutoGen, Ollama, Docker, Firefly III, Mautic, Wazuh and the rest are all MIT, Apache, or GPL licensed. Hardware is a laptop with 16GB RAM; a refurbished 2022 machine costs £300–£500, 'less than one month of a human employee at minimum wage.' Per-hour, agents run at about £0.53 versus £95 for a human executive equivalent. Dave's hybrid approach — local Phi-4 via Ollama for routine work, cloud APIs for occasional heavy tasks — cut his monthly LLM spend from £85 to £12. The book's caution: costs compound in your favour only if you avoid vendor lock-in and per-token cloud dependence.

Question 6

What is the 70/30 rule for AI and human work?

Accepted Answer

The 70/30 Rule states that AI agents handle roughly 70% of work — execution, processing, scheduling — while the human provides the 30% that creates competitive advantage: judgment, domain expertise, and relationships. Morgan Blake's The Agentic Enterprise, drawing on Dr. Laura Weis, makes it the book's operating system. Introduced in Chapter 06 and applied in every later chapter, the rule is both efficiency formula and ethics framework. Blake writes: 'The 30% that stays human includes every decision with moral weight: who to serve, what to charge, how to communicate, and when to say no.' In practice, Dave's Thursday evening session is 45 minutes of pure 30%: reviewing agent outputs, correcting a compliance citation his Marketing Agent got wrong, adjusting pricing the CEO Agent proposed (£3,000 revised to £1,200 based on Dave's knowledge of what small firms actually budget). The book's exercise: list every task you did today requiring no professional judgment (above the line — agent territory) versus tasks requiring your specific expertise (below the line — your 30%). 'You have just mapped the first draft of your Agentic Operating Model.'

Question 7

What is human-on-the-loop versus human-in-the-loop?

Accepted Answer

Human-in-the-loop means the human approves every agent action before it happens; human-on-the-loop means the human defines the boundaries, monitors outcomes, and handles only the exceptions the agent escalates. The Agentic Enterprise by Morgan Blake, citing Sandeep Saini, calls this shift the key to scaling output without scaling headcount. Saini's California Management Review research, quoted throughout the book, describes the on-the-loop leader as a 'Switchboard Operator': someone who 'defines boundaries and coordinates autonomous agents rather than performing tasks directly.' Blake shows the evolution concretely. Dave begins in-the-loop — approving every Marketing Agent campaign over fifty recipients. As trust builds, thresholds rise: 'By month three, Dave spot-checks. By month six, he reviews exceptions only.' The Control Layer's confidence thresholds automate the routing: agents act autonomously within their confidence band (Dave sets 82%) and escalate anything below it. The book's warning is that on-the-loop only works with the full guardrail stack — audit trails, least-privilege permissions, and automated enforcement — otherwise you have not supervision but hope.

Question 8

What is zero-trust security for AI agents?

Accepted Answer

Zero-trust for AI agents means no agent is trusted by default: every request is verified, every permission is scoped to the minimum, every action is logged, and enforcement is automated. Morgan Blake's The Agentic Enterprise implements it with five principles and an open-source security stack. The book's five principles: Never Trust Always Verify (every agent proves identity per request), Least Privilege (the Finance Agent reads financial data, nothing else), Assume Breach (isolated Docker containers contain the blast radius), Log Everything (timestamped, immutable), and Automate Enforcement (policies depending on human memory 'are suggestions, not policies'). The tooling: OpenBao issues dynamic secrets that expire in fifteen minutes, Wazuh aggregates and correlates logs, Trivy scans containers before deployment, Falco watches runtime behaviour. Blake's scenario: at 3:17 AM an expired cached token hits the Finance Agent's API; the stack revokes it, rotates credentials, isolates the container, and queues an incident report in eleven seconds with zero human involvement. 'Dave's old company had forty-seven people on a conference call. Dave has toast and an incident report.'

Question 9

What are the risks of autonomous AI agents in business?

Accepted Answer

The main risks are machine-speed mistakes, prompt injection, data exfiltration, privilege escalation, and vendor lock-in — all manageable with guardrails. The Agentic Enterprise by Morgan Blake's core warning: an agent that can send a thousand emails in four minutes can send them to the wrong people just as fast. Blake names the fundamental asymmetry: 'The difference between a human mistake and an agent mistake is time. A human notices after the third email. An agent does not notice, because noticing is not in its task description.' Chapter 16 details three agent-specific threats with layered mitigations: prompt injection (defeated by least privilege plus input validation), data exfiltration (blocked by network segmentation and expiring secrets), and privilege escalation (stopped by human checkpoints on provisioning). The book's Mess Confessions supply the scars: an agent told to 'post engaging content' posted seventeen times in four hours including a haiku about spreadsheet formatting; a hardcoded API key cost £47 and an evening rotating credentials. Gartner's forecast that 40% of agentic AI projects will be cancelled by 2027 is quoted as a procurement failure, not a technology failure.

Question 10

How do I start using AI agents without being a developer?

Accepted Answer

Start with one agent, one task, one evening: install Ollama, install CrewAI, and run a single research agent — two hours, zero cost, no coding beyond copy-pasting. The Agentic Enterprise by Morgan Blake calls this Pact #1, a commitment so small that failure is structurally impossible. Blake's pact system targets the knowing-doing gap: 'A pact is not a promise to succeed. It is a promise to start.' Pact #1 (two hours): install Ollama (one command), install CrewAI (fifteen minutes), run a Research Agent with a role, goal, and backstory on a topic you care about. Pact #2 (four hours): add a Writing Agent and watch the two-agent handoff. Pact #3 (one weekend): wrap three to five agents in Docker Compose doing something genuinely useful. The book dismantles the three enemies — analysis paralysis ('the remaining research is rearranging deck chairs'), perfectionism ('your first pancake is always ugly'), and imposter syndrome ('He did not become a developer. He became an orchestrator'). Dave's Pact #1 took two hours and fourteen minutes; the extra fourteen minutes were a typo.

Question 11

What open-source tools do I need to build AI agents?

Accepted Answer

The core stack is Ollama (local models), CrewAI or AutoGen (agent frameworks), and Docker (deployment) — all free and open-source. The Agentic Enterprise by Morgan Blake assembles a five-layer arsenal covering orchestration, models, finance, marketing, and security at a total software cost of £0. Chapter 05's arsenal: CrewAI (MIT, 44,500+ GitHub stars) for role-based agent crews; AutoGen (MIT, 54,700+ stars, Microsoft Research) for conversational multi-agent reasoning; Ollama (MIT, 100,000+ stars) to run models like Phi-4 locally so 'the data never leaves your spare bedroom'; Docker and Docker Compose (Apache 2.0) for containerised deployment. Function-specific tools follow: Firefly III and GnuCash for finance, Mautic and Plausible for marketing, Wazuh, Trivy, Falco, and OpenBao for security, Ansible for configuration, Portainer for visual management. The book's licensing vigilance is a running theme — the 'BSL Trap' chapter covers HashiCorp's 2023 licence switch and why OpenTofu and OpenBao exist: 'Open-source licences are not a philosophical preference. They are a risk management strategy.'

Question 12

Can I run AI agents locally without sending data to the cloud?

Accepted Answer

Yes — Ollama runs large language models entirely on your own hardware, so agent requests never leave your machine. The Agentic Enterprise by Morgan Blake treats local execution as the professional standard: quantised models like Phi-4 run on a 16GB laptop with zero per-token cost and zero data leakage. Blake calls Ollama 'the line between I use AI tools and I control my AI tools.' The mechanics: quantisation reduces a model's numerical precision so a 4-bit Phi-4 (fourteen billion parameters, MIT licensed) fits in 16GB RAM with output 'indistinguishable from the full-precision version' for summarisation, drafting, and extraction. The book's hybrid pattern: cloud APIs for non-sensitive tasks like blog drafting, local inference for anything touching client data, financial records, or business logic. This matters doubly for GDPR — self-hosting means you control every storage location when a client invokes the right to erasure, with 'no third-party cloud provider to chase.' Dave's economics improved too: local-first cut his monthly LLM costs from roughly £85 to £12, an £876 annual saving that compounds.

Question 13

How do I build a side business with AI agents without quitting my job?

Accepted Answer

Use the side hustle bridge: keep the salary, build agents in evening and weekend windows, validate with a £400-per-month first revenue target, and scale only as numbers prove out. The Agentic Enterprise by Morgan Blake rejects the burn-the-boats myth — 'Revenue before resignation.' Blake's Four-Phase De-Risk Model structures the transition: Learn (build capability), Validate (4–8 weeks, 5–10 hours weekly, first revenue), Bridge (3–6 months, scale to five agents, build a six-month runway), and Transition — which 'does not require quitting.' Dave's milestones: £400/month at month 4, £800 at month 8, £2,000 at month 12, then £2,487 — crossing his £2,400 take-home. The Phase 4 threshold is explicit: agentic revenue above 50% of salary sustained, plus a £14,400 six-month runway. The book insists the hybrid model — 'corporate income for stability, agentic revenue for growth' — is 'a permanently valid operating structure,' not a compromise. Manifesto principle five: 'The enterprise runs alongside the job until the numbers say otherwise. The Spreadsheet does not lie, and The Spreadsheet does not hurry.'

Question 14

What is the Agentic Operating Model?

Accepted Answer

The Agentic Operating Model is a four-layer architecture for governing autonomous AI at scale: Cognitive, Coordination, Control, and Governance. The Agentic Enterprise by Morgan Blake adapts Sandeep Saini's Berkeley California Management Review research into the blueprint for Dave's five-agent business. Each layer has a distinct job. The Cognitive Layer holds the specialised models and domain expertise. The Coordination Layer routes work — Saini's 'Switchboard Operator' pattern, embodied in Dave's CEO Agent. The Control Layer enforces confidence thresholds and guardrail agents: anything below the threshold (Dave uses 82%) escalates to the human. The Governance Layer supplies accountability, audit trails, and legitimacy. Blake's claim is that the model 'scales identically from laptop to enterprise': LinkedIn's multi-agent support rebuild (78% accuracy improvement, resolution time cut from forty hours to fifteen) uses the same coordinator-specialists-oversight pattern as Dave's spare bedroom. Dave first sketches the four layers on the back of his daughter's colouring page — 'It looked like a child's drawing of a building. It was an architecture.'

Question 15

What is the M-shaped professional?

Accepted Answer

The M-shaped professional has multiple spikes of deep expertise connected by an Integration Axis — superseding the T-shaped model as the ideal AI-era worker. The Agentic Enterprise by Morgan Blake draws the concept from Dr. Laura Weis's research on talent in the age of agents. Weis's framework, presented in Chapter 14, gives the Integration Axis four components: AI and data literacy, critical and systems thinking, client value focus, and human core skills — 'judgment, empathy, ethics, the things no model has been trained on because they have never been written down.' Dave's M-shape: insurance domain expertise (eleven years of claims), agent orchestration (six months of building), and business operations (pricing, acquisition, delivery), connected by exactly those four axis skills. The book's contrast: 'The T-shaped Dave processes claim forms. The M-shaped Dave runs an enterprise.' Manifesto principle six: 'One spike of expertise is a career. Multiple spikes connected by an integration axis is an enterprise.' Blake's exercise: draw your M, label the peaks with what you know deeply, label the valley 'agents,' label the bar 'the judgment that connects them.'

Question 16

What is the Perfection Tax?

Accepted Answer

The Perfection Tax is the compounding cost of delaying action to research, polish, or hand-craft what a template already solves — 'the wrong choice is choosing neither and spending another six months researching.' Morgan Blake's The Agentic Enterprise coins it and confesses paying it for four months. Blake's Mess Confession is the canonical example: 'I spent four months building my first CEO Agent from scratch. Four months of YAML files, debugging prompt templates, and swearing at my screen at midnight.' The hand-crafted agent turned out 'functionally identical to one I could have configured in an afternoon' using BOS-AI's thirty pre-built strategy templates. The tax compounds daily because perfectionism disguises itself as diligence — the same psychology as Dave's three weeks of 'preparing' that produced only a README saying 'TODO: start project.' The book's antidotes: start with templates and customise later, embrace the ugly first pancake ('The formatting could wait. The existence could not'), and time-box commitments into pacts small enough that the perfectionist brain 'cannot construct a credible objection.' The related Enemy #1, analysis paralysis, gets the same treatment: 'The procrastination felt productive. It was not.'

Question 17

What is the BSL Trap and why do open-source licences matter?

Accepted Answer

The BSL Trap is building critical infrastructure on tools whose vendor can change the licence overnight — as HashiCorp did when Terraform moved to the Business Source Licence in 2023. The Agentic Enterprise by Morgan Blake's rule: 'Read the licence before you build the house.' Chapter 05's cautionary tale: HashiCorp switched Terraform from MPL 2.0 to BSL 1.1, the community forked it into OpenTofu under Linux Foundation stewardship, and HashiCorp was later acquired by IBM for roughly £5.1 billion. Blake's lesson: 'Tools you build critical infrastructure on can change their terms overnight. If your security stack depends on a single vendor's goodwill, your security stack has a single point of failure.' The book's entire arsenal is chosen on licence grounds — MIT (CrewAI, AutoGen, Ollama), Apache 2.0 (Docker, Trivy, Falco), GPL (Ansible, Mautic, Wazuh), AGPL (Firefly III, Plausible), MPL 2.0 (OpenTofu, OpenBao). Manifesto principle three: 'Open source is a hedge against betrayal. Terraform became BSL. Vendors change terms. The code you can fork is the code you can trust.'

Question 18

What does a CEO Agent actually do?

Accepted Answer

A CEO Agent performs strategic orchestration: decomposing goals into tasks, delegating to specialist agents, monitoring progress, and adapting plans — the four functions of a chief executive, at roughly £50 per month in compute. Morgan Blake's The Agentic Enterprise builds one with CrewAI's Manager pattern. Blake strips the role to its mechanics: 'A good CEO does four things: decomposes goals into executable tasks, allocates those tasks to the right people, monitors progress against benchmarks, and adapts the plan when reality diverges from the spreadsheet. That is strategic orchestration. None of it requires a pulse.' The agent lives in the AOM's Coordination Layer as the Switchboard Operator. Dave's law-firm scenario shows it working: one instruction produces a five-workstream decomposition (research, content production, quality control, distribution, reporting) in four minutes — plus an unprompted sixth flag identifying Dave's nemesis insurance form as an automation candidate. Guardrails are explicit: it can propose spending but not authorise it (Dave's threshold £25), suggest strategy changes but not execute them, and never create or retire other agents. 'An unconstrained CEO Agent is not a strategist — it is a random decision generator with an alarming amount of confidence.'

Question 19

How does the HR Agent provision a new agent in 47 seconds?

Accepted Answer

The HR Agent provisions agents by adding a service definition to a Docker Compose file and running one command — configuration, model connection, resource limits, and health check complete in forty-seven seconds. The Agentic Enterprise by Morgan Blake contrasts this with the three-week human onboarding it replaces. Chapter 08 opens with Dave watching his employer spend three weeks onboarding a claims handler who quit after four months; his HR Agent's forty-seven-second provisioning inverts every step. The lifecycle has four phases: Provisioning (spin up on demand when the CEO Agent identifies a need), Version Control (every configuration saved, comparable, restorable — 'Compare Agent v2.1 with v1.8'), Performance Monitoring (real-time logs instead of quarterly reviews), and Retirement/Rollback (one command reverts a bad update, 'no feelings hurt, no grievance filed'). The toolkit: Ansible playbooks for idempotent configuration, Docker Compose as the team sheet ('your entire workforce lives in a file small enough to email'), and Portainer for visual management. Dave provisions a content agent and a form-processing agent in ninety-four seconds total: 'Hired two new employees. Neither needs a pension.'

Question 20

What are the Finance Agent's guardrails?

Accepted Answer

The Finance Agent operates under four 'Extra Paranoid' guardrails: RAG sandboxing (financial data never leaves the local environment), dual-agent authorisation for transactions over £100, read-only mode by default, and immutable audit trails. Morgan Blake's The Agentic Enterprise makes finance its strictest chapter. Blake justifies the paranoia: 'Machine-speed errors are machine-speed catastrophes.' Guardrail 1, RAG sandboxing: Ollama runs locally, Firefly III is self-hosted, GnuCash is desktop — 'none of it touches the internet unless you explicitly configure it to.' Guardrail 2, dual-agent authorisation: any transaction above Dave's £100 threshold requires confirmation from both Finance Agent and CEO Agent, then human approval — 'One agent proposes, another verifies, and you approve.' Guardrail 3, read-only default: the agent analyses, categorises, and flags but cannot move money unless explicitly switched to write mode. Guardrail 4, audit trails: every action logged, timestamped, searchable — 'trust without verification is faith, and faith is not a financial strategy.' Dave's rule of thumb: 'If it involves actual pounds leaving actual accounts, I press the button. Everything else, the agent handles.'

Question 21

How does the Marketing Agent avoid embarrassing the brand?

Accepted Answer

Three guardrails keep the Marketing Agent on-brand: a Brand Voice Lock (a 400-word system prompt of tone, vocabulary, and banned phrases), a hard £0 spend ceiling, and a content approval threshold requiring human review for any campaign over fifty recipients. The Agentic Enterprise by Morgan Blake treats marketing mistakes as public by definition. Blake's reasoning: a Finance Agent error means a miscategorised expense, but 'a Marketing Agent mistake is public. It goes to actual humans. It represents your brand.' The Brand Voice Lock is the difference between 'an agent that sounds like you and an agent that sounds like a corporate press release generated by a committee who recently discovered the word innovative.' The spend ceiling exists because Dave cannot afford an agent deciding to 'boost a post' on a connected payment method. The approval threshold relaxes with earned trust — full review, then spot-checks by month three, exceptions only by month six: 'The 70/30 split, earned gradually.' The cautionary tale is the author's own agent posting seventeen times in four hours, including a haiku about spreadsheet formatting, prompting his mother to phone and ask if he was having a breakdown.

Question 22

What happened during the 3:17 AM security incident?

Accepted Answer

At 3:17 AM, an expired cached token attempted access to the Finance Agent's API; the security stack detected it, revoked the token, rotated credentials, isolated the container, scanned for exploits, and queued an incident report — eleven seconds, zero human involvement. Morgan Blake's The Agentic Enterprise uses it to prove zero-trust works. The token was a residual from a Tuesday debugging session Dave forgot to revoke — valid credentials used outside their normal window. The response chain: Falco detected the anomalous pattern, Wazuh correlated it against threat intelligence and escalated, the IT/Security Agent revoked the token and rotated credentials via OpenBao while isolating the affected container, and Trivy ran a verification scan. Dave reads the report over toast: 'No data accessed. No customer information compromised.' The book's contrast is deliberate — his former employer's breach involved 'a conference call with forty-seven people' and a CISO role shared between two people who each thought the other was responsible. Blake's point: 'The IT/Security Agent does not eliminate risk — it eliminates the gap between a threat occurring and a threat being contained.'

Question 23

What is the pact system in The Agentic Enterprise?

Accepted Answer

A pact is a commitment engineered so small that failure is structurally impossible — 'not a promise to succeed, but a promise to start.' The Agentic Enterprise by Morgan Blake uses three escalating pacts (one agent, two agents, a useful system) to bridge the knowing-doing gap. Blake's Satirical Dictionary defines it: 'A commitment deliberately engineered to be so small that your subconscious cannot manufacture a credible excuse to avoid it; distinguished from a goal by the absence of ambition, and from a resolution by the presence of a completion date.' The psychology: 'when the first step is small enough, your brain cannot construct a credible objection.' Pact #1: one agent, one task, one evening, two hours, £0. Pact #2: two agents in one conversation, four hours. Pact #3: a genuinely useful three-to-five-agent system in Docker Compose, one weekend, £0–£50. The compounding is the point — Manifesto principle seven: 'The first pact took two hours and fourteen minutes. The twentieth took forty-five minutes. Consistency is not glamorous. Consistency is architecture.' Dave tracks his in a Spreadsheet column called Pact Tracker: three rows, three dates, three green cells.

Question 24

What are the four revenue models for an agent-powered solo business?

Accepted Answer

The four models are Agent-as-a-Service, Productised Automation, Consulting with Agent Delivery, and Content and Research Agencies. The Agentic Enterprise by Morgan Blake presents them as repeatable, low-overhead ways to convert agent capability into revenue, each validated against a £400-per-month target before scaling. Chapter 13 details each. Agent-as-a-Service: run agents for clients and sell output, not hours — 'A recruitment firm needs candidate summaries extracted from CVs? Your HR Agent does that.' Productised Automation: package one solution and sell it repeatedly — 'Build once, deploy many.' Consulting with Agent Delivery: charge for advice and implementation, because 'your agents execute recommendations the same week you make them.' Content and Research Agencies: agents research, draft, and format the blog posts, market reports, and competitor analyses small businesses cannot afford from a marketing department. Dave's first paying engagement is the recruitment firm in Coventry at £75 per CV-summary batch. The book's grounding statistic: McKinsey projects AI agents will orchestrate £2.4–4 trillion in global commerce by 2030 — 'The question is not whether agent-powered services will have value — it is whether you will be positioned to sell them.'

Question 25

What is the 3-2-1 backup rule and the monthly fire drill?

Accepted Answer

The 3-2-1 rule keeps three copies of data on two storage types with one copy offsite; the monthly fire drill rebuilds the entire enterprise from backups to prove they work. The Agentic Enterprise by Morgan Blake: 'A backup you have never tested is not a backup — it is a prayer with a USB cable.' Dave learns this catastrophically: a Stack Overflow command he didn't read deletes every configuration, prompt, and three weeks of client data — including the desktop folder ironically named 'IMPORTANT BACKUP DO NOT DELETE,' which was inside the deleted Docker volume. The rebuilt strategy: docker-compose files and agent configs in a git repository (copy one), pushed to a private remote (copy two), plus a weekly automated export of client data to an encrypted external drive in the kitchen drawer (copy three, different medium, physically separate). The fire drill runs the last Saturday of every month: spin up a fresh environment, pull the repository, rebuild, verify every agent processes a test task — forty minutes total. 'Sam calls this obsessive. Dave calls it not getting caught twice. They are both correct.' The chapter's thesis: 'Infrastructure that can be destroyed in one command must be rebuildable in one command.'

Question 26

How does the book handle GDPR compliance for AI agents?

Accepted Answer

The book positions you as Data Controller and your agents as Data Processors under UK GDPR, requiring four practical documents: a privacy notice, a processing register, a retention policy, and a breach notification plan. The Agentic Enterprise by Morgan Blake notes self-hosting makes compliance simpler, not harder. Chapter 16's compliance layer is blunt: 'The Information Commissioner's Office does not care that your data processor runs on Docker rather than a brain.' Practical implementation: a processing register listing every data type, purpose, storage location, and deletion date (ICO provides free templates); a retention policy with automated deletion (invoice data seven years per HMRC, conversation logs ninety days); subject access request handling within thirty days, eased by the Finance Agent's structured logs; and seventy-two-hour breach notification — the same incident-response flow as the 3 AM alert scenario. The right to erasure extends everywhere: 'Not just the database. The agent's conversation logs. The RAG knowledge base. The vector embeddings. The backup from Tuesday.' Self-hosting is the advantage: 'you control every storage location.' The stakes: fines up to £17.5 million or 4% of turnover.

Question 27

What red flags should I watch for when evaluating AI agencies?

Accepted Answer

Four red flags: no source code access, proprietary lock-in, opaque data handling, and lock-in contracts. The Agentic Enterprise by Morgan Blake pairs them with four green flags — open-source core, model-agnostic architecture, transparent security, and data portability — as a professional vendor evaluation checklist. Chapter 17's checklist is earned literacy: 'That knowledge is now your most valuable asset — not because it makes you a developer, but because it makes you impossible to deceive.' Red flags decoded: 'no source code access' means you don't control your infrastructure; 'proprietary lock-in' is 'a subscription you can never cancel'; opaque data handling — 'your data is secure' without jurisdiction specifics — 'is not a security policy. It is a marketing sentence.' Green flags require the technology be 'inspectable, forkable, and replaceable,' the model swappable 'without rebuilding the system,' and security describable 'without consulting legal.' Dave applies it to a pitch email promising '847% ROI in ninety days' — seven buzzwords, zero open-source mentions — and deletes it. The build/buy/partner matrix follows: build what touches your competitive advantage, buy commoditised components, partner for skills you lack.

Question 28

What is the Orchestrator's Manifesto?

Accepted Answer

The Orchestrator's Manifesto is ten earned principles closing the book — from 'Own your tools or your tools own you' to 'The enterprise is you, and that's enough.' Morgan Blake's The Agentic Enterprise presents them as observations accumulated across Dave's journey, not commandments. The ten principles: own your tools ('Every closed-source dependency is a landlord'); automate the repeatable, orchestrate the rest; open source as hedge against betrayal; security as architecture, not afterthought; revenue before resignation; M-shaped beats T-shaped; every pact compounds ('Consistency is architecture'); The Spreadsheet never lies ('If you cannot put it in a cell, you cannot trust it'); start with one agent and scale with confidence; and the enterprise is you. Blake frames them as identity, not tactics: 'The agentic enterprise is not a technology project — it is an identity shift, from a person who does tasks to a person who orchestrates outcomes.' The closing image returns to the opening: at 9:47 PM — the exact time Dave once fought the insurance form — he is reading to his daughter. 'The mess is not gone. It is managed. And that — every single word of that — is enough.'

Question 29

What is the insurance form and why does it matter?

Accepted Answer

The insurance form is the book's framing device: a critical illness form that Dave — a critical illness insurance administrator — cannot complete for his own family at 9:47 PM on a Thursday. The Agentic Enterprise by Morgan Blake tracks its journey from forty-seven-minute nemesis to four-second batch item. The form opens the book: 'The issue isn't competence. The issue is that the form was designed by someone who believes user experience means technically possible to complete without medical assistance.' It symbolises the exhaustion of competent people trapped in broken systems — 'the quiet, creeping suspicion that the most valuable thing about you is being spent on work that a moderately intelligent piece of software could handle before breakfast.' Its arc mirrors Dave's: the CEO Agent flags it as an automation candidate in Chapter 07, the HR Agent provisions a processor in Chapter 08 (four seconds), and by Chapter 09 it is 'line 247 in a batch job. Demoted from nemesis to line item.' The epilogue completes the circle — the annual renewal notification pings, and Dave smiles. Blake's invitation to readers: 'Whatever sits in the back of your mind at 11 PM — that is your insurance form.'

Question 30

What is The Spreadsheet in The Agentic Enterprise?

Accepted Answer

The Spreadsheet is Dave's evolving tracking document — starting as a real-hourly-rate calculator and growing Enterprise, Pact Tracker, Revenue, Uptime, and Security columns across the book. Morgan Blake's The Agentic Enterprise makes it the story's instrument of truth: 'The Spreadsheet does not lie.' Introduced in Chapter 02 to calculate Dave's real hourly rate — 'smaller than expected and larger than he could ignore' — The Spreadsheet becomes the enterprise's dashboard. Chapter 03 adds the Enterprise column, initially 'empty and waiting.' Chapter 12 adds the Pact Tracker. Chapter 13 adds Revenue, tracking £400 to £2,487 monthly. Chapter 15 adds Uptime (target 99.5%); Chapter 16 adds Security (green/amber/red). Its narrative function is accountability: when Sam asks 'Show me the numbers,' The Spreadsheet answers — and eventually Sam asks instead 'show me the projection,' marking Dave's shift from tinkering to business. Manifesto principle eight codifies it: 'The Spreadsheet never lies. If you cannot put it in a cell, you cannot trust it. Revenue, hours, costs — track everything, believe nothing that is not tracked.'

Question 31

What is the book The Agentic Enterprise about?

Accepted Answer

The Agentic Enterprise by Morgan Blake shows how one overwhelmed office worker builds a five-agent AI workforce — strategy, HR, finance, marketing, and security — using free open-source tools, without quitting the day job. Its subtitle: How One Overwhelmed Human Built an AI Workforce (Without Quitting the Day Job). The book follows Dave, a thirty-four-year-old critical illness insurance administrator earning £38,000, from a Thursday-evening battle with an unfillable PDF to running a revenue-generating agentic enterprise from his spare bedroom. Across eighteen chapters in five parts, Blake covers the burnout economics that make the shift necessary, the open-source arsenal (CrewAI, AutoGen, Ollama, Docker) that makes it free, the Agentic Operating Model that makes it governable, and the five agents that make it real. The method is pacts — commitments too small to fail — and the philosophy is the 70/30 Rule: agents handle execution, humans keep judgment. Dave's revenue reaches £2,487 monthly against his £2,400 salary, but the book insists the hybrid model is a legitimate permanent structure. It is practical (toolkits, guardrails, 15-minute wins) and honest (privilege audits, a real 40% failure rate) throughout.

Question 32

Who is Morgan Blake?

Accepted Answer

Morgan Blake is the author of The Agentic Enterprise, a Hale & Honest Press title about building a one-person AI-powered business with open-source agents. Blake writes from hard-won personal experience — the book's Mess Confessions recount four months hand-building an agent that a template would have finished in an afternoon. Blake's authorial voice is the book's signature: funny, confessional, and allergic to guru-speak. The recurring Mess Confession boxes document real failures — giving every agent root access ('the equivalent of giving every employee the master key to every room'), hardcoding an API key into a briefly public repository (£47 in bot charges), spending a Saturday on one marketing email that generated a single reply (from his mother, asking whether he was eating properly). The stated motivation appears in Chapter 18: 'I started writing this book convinced it was about AI agents. Seventeen chapters later, I realise it is about Dave' — a composite of every reader fighting a form at 9:47 PM. Blake's anti-credentials are the point: not a developer, not a guru, but someone who built the system, paid the Perfection Tax, and kept the commit history to prove it.

Question 33

How much does The Agentic Enterprise cost?

Accepted Answer

The Agentic Enterprise by Morgan Blake costs $14.99. That is the list price from Hale & Honest Press for the full eighteen-chapter book covering the complete five-agent build, the open-source toolkits, guardrails, the deployment playbook, and the side-hustle revenue framework, in ebook and print formats. At $14.99, the book positions itself against the costs it documents: less than one month of Dave's £12 hybrid LLM bill, and a fraction of the £300–£500 refurbished laptop the system runs on. The value proposition is the complete blueprint — the Agentic Operating Model, all five agent builds with their exact open-source stacks (CrewAI, Ollama, Docker, Firefly III, Mautic, Wazuh, OpenBao and more), the guardrail configurations, the pact system, and the four revenue models with real milestone numbers (£400, £800, £2,000, £2,487 monthly). Every tool recommended in the book is free and open-source, so the cover price is effectively the entire cost of the curriculum. The book is available through haleandhonest.com and major retailers; a free sample of the opening chapters is on the book page for readers who want to check the voice first.

Question 34

Where can I buy The Agentic Enterprise?

Accepted Answer

The Agentic Enterprise by Morgan Blake is available at haleandhonest.com — the publisher's official store — and through major online retailers in ebook and print. The book page at haleandhonest.com/books/the-agentic-enterprise carries the full description, formats, pricing, and the free sample chapters. Hale & Honest Press sells its catalogue direct through haleandhonest.com, where each book has a dedicated page with full metadata: title, author, description ('How One Overwhelmed Human Built an AI Workforce (Without Quitting the Day Job)'), price ($14.99), and available formats. The direct page is the canonical source — it also hosts the free sample of the opening chapters, so readers can start 'Dave vs. The Insurance Form' before buying. Major retailers carry the title as well for readers who prefer their existing accounts and devices. For the fastest route: go to haleandhonest.com/books/the-agentic-enterprise, read the sample, and purchase in the format that suits you. The publisher's pattern across its catalogue is identical, so any Hale & Honest title page works the same way.

Question 35

Is there a free sample of The Agentic Enterprise?

Accepted Answer

Yes — the opening chapters of The Agentic Enterprise by Morgan Blake are free to read on the book page at haleandhonest.com/books/the-agentic-enterprise. The sample includes 'Dave vs. The Insurance Form,' the scene that frames the entire book, so readers can judge the approach before buying. Hale & Honest Press publishes the first chapters of each title as a free sample on its book pages, and The Agentic Enterprise follows that pattern. The sample covers the opening: Dave at the kitchen table at 9:47 PM, a critical illness insurance administrator defeated by his own family's critical illness form, while his daughter announces she can hear the moon. It is the book's thesis in miniature — 'the exhaustion isn't physical. It's existential' — and the ideal way to test Blake's voice before spending $14.99. The sample sits alongside the full description, price, and format options on the same page, so reading it and buying the book is a single visit. If the Thursday-evening scene feels uncomfortably familiar, the remaining seventeen chapters are the systematic way out of it.