Agents: What Are They and What Are Some Example Use Cases?
A practical guide to what AI agents are, how they differ from chatbots and where finance, procurement and sustainability teams can apply them for real ROI.
Who this is for
This post is not for engineers configuring infrastructure and deployment environments. That is important work, and something Sumday supports teams to do. This guide is for people who are curious about how agents will actually be used and the role they will have in shaping them.
Intro
Agents seemed to appear out of nowhere. Not long ago, "agents" were for Hollywood stars. In the next 12 months, it would not be surprising if you had hundreds of agents working for you. That is already the case at Sumday and many fast-moving startups and corporates.
If you are just getting into planning or building, this guide shares lessons from a team that has already seen around the corner.
What is an agent?
An agent is like a virtual collaborator that can figure out how to get something done without you directing every step.
This is an agent task: "Review all invoices every 3 months and identify spend patterns that suggest people are paying for personal software subscriptions instead of an enterprise plan."
This is not an agent task: "Review this invoice. Go to this line. Google this name. Check the seat price. Go to the next invoice..."
You give an agent a goal. It decides the path, uses the tools it needs (some off-the-shelf, some custom), and keeps going until the job is done.
It works whenever you need it to: on demand, on a schedule, or continuously.
A chatbot answers questions and follows direct commands. An agent acts. It does things, makes decisions, and follows the guardrails in its instructions.
How do you build an agent?
Start with safe foundations so people outside IT can build agents responsibly. If you are unsure how to set this up for finance, procurement, and other leaders, Sumday can help.
For guidance, see the National Cyber Security Centre report released on 1 May 2026 here.
You want to ensure your leaders know how to build basic agents. This should be a priority. If it is not something people are confident in right now, book a training session with us - it may be the highest ROI thing you do this month.
You can begin right now: open Claude or Copilot and ask, "How do I build an agent if I have never done this before?"
If you want to read more: Microsoft's agent builder. Claude.
What are some agent use cases?
Below is a practical shortlist for sustainability, finance, and procurement teams. In AI transformation programs, Sumday typically helps teams stack-rank highest-ROI use cases and then build them quickly.
Finance and FP&A
Use Case | What the Agent Does | ROI Potential |
|---|---|---|
Business case review agent | Applies your methodology (discount rate, ROI thresholds, strategic criteria) to stress-test proposals before they reach finance. | High |
Variance analysis assistant | Pulls actuals vs budget, identifies material variances, and drafts commentary for review. | High |
Board and executive report drafting | Synthesises financial data into narrative-ready management reports. | High |
Month-end close checklist agent | Tracks completion of close tasks, flags blockers, and chases outstanding items. | Medium |
Budget submission review | Checks submissions for completeness and methodology consistency before consolidation. | Medium |
Scenario modelling assistant | Runs sensitivity analysis across key variables (rates, volumes, FX) and summarises outputs. | High |
Grant and funding opportunity scan | Monitors relevant funding sources and surfaces matches to current and planned projects. | Medium |
Policy and compliance Q&A agent | Answers staff questions against finance policies, delegations of authority, and procedures. | Medium |
Invoice and contract extraction | OCR extraction of key terms, values, and dates from invoices and contracts into structured data. | High |
Cashflow optimisation agent | Provides notifications of past due receivables, automates reminder emails, and supports credit monitoring. | Medium |
Treasury management agent | Produces automated FX and interest-rate forecast briefings, TD rate comparisons, and treasury policy automations. | Medium |
New starter finance onboarding | Answers common process questions, links templates, and explains business case and approval workflows. | Low-Medium |
Why we always prioritise the business case
At Sumday, we believe AI should help organisations make better, faster decisions. That often starts by directing resources to the initiatives with the best return and strongest strategic alignment.
The challenge is simple: many teams want to spend, but very few people have the broader context and skills to test assumptions and produce decision-ready business cases.
When teams cannot filter and shape ideas early, organisations see: cultural frustration, wasted finance capacity, delayed execution, weaker commercial decisions, and ESG left as an afterthought.
Procurement
Use Case | What the Agent Does | ROI Potential |
|---|---|---|
Scope of work drafting agent | Generates first-draft SOWs from a project brief using your best historical templates. | High |
Supplier evaluation assistant | Scores responses against weighted criteria and produces structured evaluation summaries. | High |
Contract review and redline assistant | Flags non-standard clauses, missing provisions, and risk areas against preferred positions. | High |
Procurement policy Q&A | Answers staff questions on thresholds, approvals, and process without pulling in procurement each time. | Medium |
Tender document summariser | Extracts key requirements, deadlines, and evaluation criteria from long tender packs. | Medium |
Spend analysis and categorisation | Classifies unstructured spend for reporting and strategic sourcing decisions. | High |
Supplier risk monitoring | Monitors news and public data on key suppliers and flags material developments. | Medium |
Purchase order and approval routing | Automates low-complexity purchase requests through the right approval chains. | High |
Market price benchmarking | Pulls and summarises market pricing data for common categories to support negotiations. | Medium |
Post-contract performance review | Synthesises delivery data and stakeholder feedback into structured supplier performance summaries. | Medium |
Sustainability and ESG
Use Case | What the Agent Does | ROI Potential |
|---|---|---|
Scope 1 and 2 data extraction | Extracts consumption data from energy bills and meter reads using OCR and maps to emissions factors. | High |
Scope 3 spend-based estimation | Applies emissions factors to procurement spend categories for Scope 3 estimation. | High |
ASRS/AASB S2 disclosure drafting | Drafts climate disclosure narrative sections against your data and the standard requirements. | High |
Emissions variance analysis | Identifies material changes in emissions by source, period, and business unit. | Medium |
Modern Slavery data collection | Ensures suppliers comply with modern slavery requirements at onboarding and on an ongoing basis through automated surveys aligned with legislation. | Medium |
Supplier ESG data collection | Automates outreach and collation of supplier emissions and ESG data for Scope 3 reporting. | Medium |
Sustainability report drafting | Synthesises performance data into structured report sections for review. | Medium |
Regulatory change monitoring | Monitors ASRS, ISSB, and related updates and surfaces relevant changes. | Medium |
Grant and green finance scanning | Identifies applicable sustainability grants, green bonds, and incentive programs. | Medium |
Materiality assessment support | Collates and summarises stakeholder input for double materiality assessments. | Medium |
Carbon reduction initiative tracker | Monitors progress against reduction targets and flags initiatives that are off-track. | Medium |
Key points
There is a lot of low-hanging fruit. Start tomorrow.
Upskill the leaders in your teams. They learn once, apply it fifty times, and lift everyone around them. That is how people-first AI transformation compounds.
Beyond the basics, think in systems: multiple agents pulling from many sources, working together to deliver complex work quickly. You want projects scoped, business cases built, and experiments running within 90 days. We help teams get there.
