Module 2 · Day 2 · AnyCompany Leader Workshop

Exploring Agentic AI

What makes a system "agentic" — the loop that drives it, the four shapes agents come in, and which shape fits which finance problem.

Day 2: Foundations ⏱ 35 min 🎯 4 Sections

The Agentic Loop — Perception → Reasoning → Action

Strip away the technology jargon and an agent is just three capabilities running in a continuous cycle. See what's happening, think about what to do, act, then look again. The loop is what separates an agent from a chatbot — and what makes the leadership conversation about where to put a human, not whether to use AI.

AGENTIC LOOP continuous cycle 👁️ Perception Action 🧠 Reasoning OBSERVE DECIDE EXECUTE
👁️

Perception

Observe. Take in text, PDFs, screenshots, API responses. For finance: read the invoice, see the dashboard, fetch the bank statement.

🧠

Reasoning

Decide. Match observation against goal. Choose the next step. For finance: "this is a duplicate vendor — flag for review" or "GST applies, calculate at 9%".

Action

Execute. Call an API, post a journal entry, send the confirmation, queue for human approval. The loop continues until the goal is met.

Why The Loop Matters For Leaders

The loop is your governance hook. Each step — perceive, reason, act — is a place you can put oversight, controls, and audit trail. The L1/L2/L3 maturity model from Module 7 (Day 1) lives here:

🛡️

L1 — Human in the loop

Agent runs perception + reasoning, surfaces the proposal. Human approves every action. Audit-friendly default.

🚦

L2 — Bounded autonomy

Agent acts within preset thresholds (e.g., auto-pay invoices < SGD 25K with clean POs). Anything outside escalates.

🚀

L3 — Goal-driven

Agent owns the outcome and chooses its own steps. Reserved for low-risk, well-instrumented domains. Most finance teams stay at L1 or L2 for years.

The Four Shapes Agents Come In

Not every agent looks the same. Pick the shape that matches the problem — the wrong shape is the most common reason an agentic pilot fails. Click each card to see when it fits, what it costs to build, and a finance example.

📋

Workflow

Fixed path
🤖

Autonomous

Self-directed
🧬

Hybrid

Router decides
👥

Multi-Agent

Team of agents

📋 Workflow Agent — The Recipe

Follows a predetermined sequence — like a recipe card. The path is fixed at design time: Trigger → Step 1 → Step 2 → Step 3 → Done. The agent can call tools and reference knowledge bases at each step, but only in the prescribed order. No improvisation.

Strengths Fast. Predictable. Cheap (no extra reasoning calls). Audit-friendly — every run looks the same. Easy to test and instrument.
Limitations Brittle. Anything unexpected breaks the path. Doesn't learn or adapt. Need to update the recipe whenever the process changes.
AnyCompany finance fit: Period-close journal posting (extract from cube → validate → format → post → notify), AP three-way match, vendor onboarding KYC checks. Anything where the steps are written down already.

Pick the Shape — Decision Table

Dimension📋 Workflow🤖 Autonomous🧬 Hybrid👥 Multi-Agent
Path planningFixed sequenceAgent plans each stepRouter picks workflow vs autonomousCoordinator + specialists
Speed⚡ Fastest🐢 Slower (reasoning loop)⚡ Balanced🐢 Variable
Predictability✅ High⚠️ Lower✅ Medium⚠️ Variable
Cost per run$ Cheap$$$ Higher (more LLM calls)$$ Mid$$$$ Highest
Build effortLowMediumHighVery high
Audit trailTrivialNeeds structured loggingPer-branch loggingCross-agent tracing
Finance use casePeriod-close journalsVendor escalation triageTax helpdesk + escalationQuarter-end orchestration
📈 The leadership read: Most finance teams should start with workflow agents for 70–80% of work, add a hybrid agent as the second pilot, and only consider multi-agent systems once the first two are in production with measurable impact. The temptation is to design the multi-agent dream-system first. Resist it.

Tools & Memory — The Two Things Agents Reach For

An agent without tools is just a chatbot. An agent without memory forgets every step it took. Tools let the agent do; memory lets it continue. This module is the conceptual intro — the practical setup is in M5 Skills, M6 Plugins/Connectors, and M7 Cheat Sheet.

What's a Tool?

A tool is anything the agent can call to interact with the real world. Your existing systems — ERP queries, bank rail status checks, the policy library, the chargeback ledger — become tools the moment you wrap them and tell the agent they exist.

🗄️System queries 📁File operations 🔍Knowledge-base search 🌐Web / API calls 🧮Calculations & transforms 📨Notifications & updates

Finance-Shape Tools — A Concrete Tool Set

What a vendor-escalation agent might be given access to:

ToolWhat it doesWho owns it
get_invoice(id)Pull invoice header + line items from ERPFinance Systems / IT
check_payment_status(invoice_id)Status from the bank rail / payment gatewayTreasury
match_po(invoice_id)Three-way match against PO + receiptProcurement
search_policy(query)RAG over your tax / treasury / accounting policy libraryKnowledge owner per domain
post_journal(entry)Post a journal entry with full audit trailControllership (high-risk — usually behind L1 review)
notify_vendor(vendor_id, message)Send vendor portal / email updateProcurement / AP

Memory — Two Kinds That Matter

💭

Working memory

What the agent has done in this run. The reasoning chain so far. Stays alive for the duration of the task — minutes, not days. Drives "what's the next step?".

🗄️

Long-term memory

What the agent has learned over time — vendor preferences, your team's escalation patterns, common exception cases. Survives across runs. This is where the productivity compounding comes from.

Standardising How Agents Reach Tools — MCP

Every framework used to invent its own way to wire up tools. Model Context Protocol (MCP) is the emerging standard — a universal protocol so any agent (Claude Cowork, Bedrock AgentCore, etc.) can use any compliant tool. Your IT team writes the tool once as an MCP server, and any agent your business uses can call it.

👉 Module 6 (Plugins, Connectors & MCP) covers the operational setup — install, activate, authenticate — and includes a live demo of the AWS MCP connector pulling Bedrock pricing data into Cowork in real time.

Pattern-Matching Real Finance Work to Agent Shapes

Take the eight functions in the room. Match each to the agent shape that best fits. This is the workshop's most useful slide — print it, take it back to your team, use it as the starting point for your first build.

📋Procurement — Contract Clause Reviewer
Read incoming MSA → compare against standard template → flag deviations → draft redline summary. Linear, repeatable.
WorkflowQuick win
📞Audit / Integrity — Whistleblower Triage
Read submission → classify severity → look up similar prior cases → propose investigator routing. Human signs the routing decision.
HybridSensitive
📊FP&A — Variance Commentary Writer
Pull actuals + forecast + last quarter's commentary → categorise drivers → draft narrative → flag the three meeting-worthy items.
HybridFlagship
📑Reporting — Disclosure Note First-Drafter
From this period's movement schedules → IFRS / local-GAAP notes consistent with prior periods → ready for the human signer.
WorkflowFlagship
🧾Tax — Internal Helpdesk
"Is this expense GST-claimable?" / "withholding rate for SG → VN service payment?" → answers from your tax policy library, with citations.
HybridRAG-grounded
🏦Controllership — Quarter-End Orchestrator
Multiple specialised agents (journal validator, intercompany recon, disclosure drafter, variance commentary) coordinated by a controller-coordinator agent.
Multi-AgentAspirational
💰Treasury — Weekly Cash Narrative
Pull cash positions → draft weekly narrative → highlight covenant-relevant moves. Same shape every week.
WorkflowQuick win
🗂️Data & Analytics — Ad-Hoc Query Translator
Finance user asks question → agent writes SQL → runs it → narrates result. Open-ended; agent decides which tables to join.
HybridFlagship

Agent-Shape Heuristic for Your Team

If your process looks like…Pick this shape
Same steps every run, well-documented, you could SOP it📋 Workflow agent
Steps depend on what the agent finds (anomaly investigation, vendor query)🤖 Autonomous agent
Most cases are routine, but some need open-ended investigation🧬 Hybrid agent (router + workflow + autonomous)
Multi-domain process spanning several specialised teams (close, audit, group reporting)👥 Multi-agent system — but build the single agents first, orchestrate later
🎯 Coming up: Module 3 — Workflow Patterns — zooms into the four ways agents chain their reasoning steps inside the workflow shape. Then Module 4 introduces the Cowork stack (Project Instructions / Skills / Scheduled Tasks / Plugins) — the operational vocabulary you'll use for the build exercise.