Strategic Analyst
Reads KPI dashboards, surfaces variances and anomalies, drafts a corrective-action memo for leadership review.
Framework map · 14 named agents · 3 tiers
AI across three organisational levels.
Every business operates on three layers, leadership, operations, execution. An AI workforce should too. This page maps the 14 agent types we build, organised by the level of the business they fit. Use it to decide where AI should enter your business first.
What are the three organisational levels for AI?
theagency47 organises AI agents into three levels that mirror how businesses already operate: Executive (management layer, decision-support and briefings, low volume, high judgment), Operational (middle layer, departmental flows like sales, support, finance, medium volume, KPI-measurable), and Task (execution layer, repetitive bounded work like data entry, email triage, document formatting, high volume, binary-correct). Every business has all three layers; an AI workforce should cover all three to avoid leaving high-leverage work on the table. We have 14 named agent types in production-ready templates across these tiers.
theagency47 · Updated May 2026Why the level matters before the agent
The most common AI agent failure is putting the wrong kind of agent on the wrong kind of work. A €15K custom agent built to "answer FAQ" is over-engineered. A SaaS chatbot drafted to write a monthly board briefing is under-powered. Both fail predictably, and both could have been avoided by deciding the level first.
Every business has three levels of work going on simultaneously:
- Leadership work, decisions, briefings, scanning for what matters. Low volume, high judgment, expensive humans.
- Operational work, running the department day-to-day. Medium volume, KPI-measurable, where most of the team's time goes.
- Task work, repetitive bounded actions that nobody loves. High volume, low judgment, where the team's morale leaks.
An AI workforce should cover all three. The depth and order in which it covers them is what we map below, and what a discovery call (or a written agent brief) figures out for your specific business.
Level 1 · Management layer · Executive agents
AI that supports leadership decisions.
Senior-level work: read across many sources, surface what matters, produce briefings a human with judgment can act on. Low volume (a few outputs per week), high judgment, almost always built on Anthropic Claude Opus.
Board Briefing Agent
Pulls from multiple sources weekly: CRM, financials, product analytics, market signals: and drafts the discussion document for board / partner meetings.
Market Intelligence Agent
Monitors competitors, market shifts, customer-relevant news. Produces a weekly digest with what changed and what it means.
HR Strategist
Evaluates CVs against role spec, prepares interview guides, structures onboarding plans. Senior-level support for the talent function.
Best for: CEOs, founders, partners. The first time an Executive-tier agent feels valuable is the first Monday morning a leader opens a draft briefing instead of starting from a blank document. Typical scope: 1–2 agents at this tier per engagement.
Level 2 · Middle layer · Operational agents
AI that runs your departmental flows.
The bulk of where teams spend time. Recurring departmental processes, sales outreach, customer support, content production, project management, finance. Medium volume, KPI-measurable, Claude Sonnet default. Most clients start here.
Sales Outreach Agent
Researches prospects, drafts personalised cold emails in your voice, sequences follow-ups, triages replies, books meetings.
Customer Support Tier-1
Resolves routine support tickets (order status, FAQ, basic troubleshooting, bounded refunds), escalates the rest with full context.
Content Production Agent
Drafts blog posts, social media, internal communications: all in your brand voice, with internal-link discipline and SEO structure.
Project Manager Agent
Tracks task status across tools, drafts weekly status reports, identifies blockers, prompts owners on overdue items.
Finance Agent
Categorises expenses, processes invoices, flags cash-flow anomalies, drafts month-end summaries for review.
Best for: department heads, operations leads. Most ROI math breaks even fastest here because the time savings are concrete and the KPIs are already measured. Two of our three showcase agents, Sofia and Yiannis, live at this tier.
Level 3 · Execution layer · Task agents
AI that absorbs the work no one loves.
High volume, bounded inputs, well-defined outputs. The work that drains team morale and never appears on a client invoice. Claude Haiku usually fits, fast, cheap per call, accurate enough for bounded classification.
Data Entry Agent
Reads structured and semi-structured data from PDFs, emails, forms: writes records to CRM, spreadsheet, or database. High volume, low cognitive load.
Email Triage Agent
Watches inbox, categorises by urgency and topic, drafts routine replies, surfaces only what needs human attention. Daily digest format.
Document Formatter
Takes drafts in any format and produces reports, contracts, presentations matching your firm's style guide. Removes the unloved last 20% of polishing.
Meeting Notes Agent
Transcribes meetings, summarises key points and decisions, extracts action items with owners and dates, drops them into your task tracker.
Translation / Localisation Agent
Translates between EL ↔ EN (or other language pairs) with consistent terminology across documents, brand voice preserved, glossary-enforced.
Best for: any team with high-volume repetitive work. The morale impact often matches the productivity impact, junior team members get to do interesting work instead of the parts they hate. Our third showcase agent, Maria, lives here.
The agency advantage: a fleet, not a single agent
theagency47's core argument is in the framework above: we don't deliver "an AI agent." We deliver an AI fleet, a coordinated set of agents across levels, with shared knowledge bases, defined handoffs, and unified KPIs. Buying one operational-tier agent leaves the executive and task work uncovered. Buying generic AI tooling without a tier model leaves you with five copies of the same shape of agent. The fleet model is what makes the math work.
A typical engagement shape (see services for full breakdown):
| Engagement | Level 1 | Level 2 | Level 3 | Total |
|---|---|---|---|---|
| Spark · €2,500 | — | — | 1 agent | 1 agent |
| Workforce Starter · €7,500 | 1 agent | 1 agent | 1 agent | 3 agents |
| Workforce Pro · €15,000 | 1 agent | 2–3 agents | 1–2 agents | 5 agents |
| Enterprise · €20,000+ | 2–4 agents | 4–8 agents | 2–5 agents | 8–15+ agents |
Where most clients start
Three patterns we see across engagements:
- Operational-first (most common). Pick the one departmental flow that costs you the most unbillable hours per week, usually sales outreach or customer support. Build an Operational agent at Tier 2 with a Task agent at Tier 3 supporting it. Once it ships and is measurable, expand.
- Task-first (entry-tier engagements). Lowest-risk way to test if AI agents fit your workflow at all. A single Spark engagement with a Tier-3 task agent, 14 days, €2,500. If it works, scale up.
- Department-wide (most leverage). A Workforce Pro covering all three tiers in one department, typically customer success, sales operations, or finance. 60 days, €15,000. Highest unit-of-work impact.
The wrong starting point is "Executive-tier first", almost always. Without operational data already flowing, an Executive agent has nothing to read.
FAQ
Questions about the levels.
What are the three organisational levels for AI agents?
Executive (management layer, decision-support, briefings, low volume, high judgment), Operational (middle layer, departmental flows, medium volume, KPI-measurable), Task (execution layer, repetitive bounded work, high volume, binary-correct).
Which level should a business start with?
Almost always Tier 2 or Tier 3. Sales outreach (Sofia) and email triage (Maria) are the most common entry points. Executive-tier agents are higher-leverage but typically the second or third agent deployed, not the first.
Can a business deploy all three levels at once?
Yes. Workforce Pro (5 agents, 60 days, €15,000) covers all three tiers in a department. Workforce Starter (3 agents, 30 days, €7,500) deploys one agent at each tier as a balanced introduction.
What about HR, marketing, or finance, where do those fit?
They span multiple tiers. HR has Executive (HR Strategist for talent decisions), Operational (onboarding orchestrator), and Task agents (CV screening). Marketing similarly spans all three. The tier framework is about the shape of the work, not the name of the department.
How is this different from the 3-tier AI workforce model blog post?
The blog post explains the theory, why the framework exists and what failures it prevents. This page is the practical operational map, which specific agents live at each tier and where most clients start. The blog is the "why"; this page is the "what fits where".