You do not need to become a software engineer to stay valuable in finance. You do need enough technical confidence to understand where your numbers come from, how systems exchange them and what an AI-generated answer has actually done.
In 2026, the useful question is no longer “Which AI tool is everyone using?” It is “Which part of my work could become more reliable, more useful or easier to review?” Start there. A finance professional who understands a process and can improve it has something much stronger than a long list of subscriptions.
This is a practical learning map for accountants, controllers, FP&A teams and business owners. You do not need all these products. Pick one assistant, keep your spreadsheet fundamentals and build one small workflow. Features and access were checked on 24 October 2026; plans, regions and administrator settings can affect availability.
Your starting stack: learn in this order
- First: Excel or Google Sheets, clean data, formula checks and one AI assistant.
- Next: one automation platform, a simple trigger and an approval step.
- Then: APIs, structured data and a small database when a spreadsheet stops being enough.
- Later: retrieval systems, cloud infrastructure and production monitoring, with technical support where needed.
1. ChatGPT: your first analysis and explanation partner
Use ChatGPT to explore a spreadsheet, explain a formula, structure a management commentary or investigate a business question with sources. File-based work, web research and connected tools can bring analysis and drafting into one conversation; availability varies by account.
Try this: upload an anonymised monthly expense CSV. Ask it to identify unusual movements, show the calculations and distinguish observations from possible explanations. Then reconcile its totals to your control total. Your knowledge of the business is what turns a plausible explanation into a useful one.
Learn first: define the period, currency, output format and evidence required. Ask for assumptions and calculation steps. A polished paragraph is not evidence.
Source: OpenAI: ChatGPT on the web.
2. Claude, Claude for Excel and Claude Code
These are related experiences with different jobs. Claude can help reason through documents and draft explanations. Claude for Excel works inside a workbook: its documented capabilities include cell-level references, formula debugging, changes to assumptions and model building. It is available on supported paid plans.
Claude Code is a coding agent that can read project files, edit code and run commands. A finance use case might be building a repeatable CSV-cleaning script or a small internal reporting tool. It is more powerful than asking a chat to explain a formula, and needs more care about files, commands and system access.
Try this: on a copy of a budget model, ask Claude for Excel to explain the cash-flow drivers and trace the cells. Review any changes before accepting them. Later, ask Claude Code to build and test a script on synthetic transactions, without production credentials.
Source: Claude for Excel documentation.
Source: Claude Code overview.
3. n8n: connect a process, including its exceptions
n8n lets you connect applications and add branching, transformations, API requests and AI steps. For finance, think of a workflow that watches an approved folder, extracts invoice fields, checks duplicates and sends exceptions to a reviewer.
Why learn it: it helps you see the process as steps with inputs, outputs and failure paths. First build: copy a new test-file name and its link into a tracking sheet. Add extraction only after that works. Managed hosting and self-hosting are different operating choices; self-hosting brings updates, backups and security responsibilities.
Source: n8n Google Drive integration.
Source: n8n hosting documentation.
4. Make: visual workflows you can inspect
Make uses visual scenarios to connect applications. It is useful when you want to see how a record moves through filters, branches and transformations. A finance team could route a submitted expense by department, assemble an approval pack or update an invoice tracker.
First build: form submission → validation → tracking sheet → review notification. Learn filters, error handling and what happens when the same input arrives twice. Check that the specific connector supports the action you need before choosing a plan.
Source: Make integration directory.
5. Zapier: a straightforward starting point for app-to-app work
Zapier connects application events to subsequent actions. A simple finance-administration example is creating a follow-up task when an approved customer-onboarding form arrives, then recording the source link.
Why learn it: a small linear workflow can teach triggers, actions and field mapping without a large build. First build: incoming approved form → task for the accounts team. Compare actual actions, limits and costs with Make or n8n on your process. There is no need to learn all three at once.
Source: Zapier integration directory.
6. Notion: turn individual knowledge into a team process
Notion combines documents and databases, with AI capabilities depending on the plan. Use it to organise close checklists, policy notes, automation runbooks and a register of who owns each control.
First build: a month-end page with owners, deadlines, evidence links and a short explanation of every exception. Give each policy an owner and review date. Keep the accounting system as the book of record; a helpful workspace should not become an uncontrolled second ledger.
Source: Notion AI and workspace capabilities.
7. Supabase: when your project needs a real database
Supabase provides a Postgres database alongside authentication, storage and other backend services. It becomes relevant when a workflow needs persistent records, controlled access or a dependable history beyond a shared spreadsheet.
Finance application: an exception queue that stores document IDs, review status and timestamps. Learn first: tables, primary keys, basic SQL and access rules. This is a later-stage tool for many beginners. Never put a privileged service key in a public page or assume that an online table is private by default.
Source: Supabase documentation.
8. Google Sheets and Apps Script: a useful first workshop
Google Sheets is accessible for collaborative trackers and lightweight prototypes. Apps Script adds JavaScript-based automation, custom functions and connections to services such as Drive and Gmail. Use it to produce a weekly status summary or validate a controlled import file.
First build: read a small test sheet, flag missing invoice numbers and prepare a summary in another tab. Learn ranges, arrays, triggers and permissions. Apps Script has quotas and execution limits, so a successful small script is not proof that a high-volume production workflow will keep running.
Source: Google Apps Script for Sheets.
Source: Apps Script quotas.
9. Keep Excel, Power Query and Power BI in the picture
Do not abandon the tools your team already relies on. Excel is still a useful modelling environment. Power Query teaches repeatable data preparation, while Power BI is a natural next step for governed reporting. Our suggested learning project is a monthly actual-versus-budget pack that refreshes from a consistent folder structure, with totals reconciled before commentary is written.
Build a first finance portfolio project with Excel, Power Query and Power BI.
10. API connections: understand the doorway between systems
An API is a defined way for software to request data or actions from another system. You do not have to write an integration immediately. Start by learning what an endpoint, authentication scope, JSON response, rate limit and error message mean.
Imagine asking an accounting system for yesterday’s unpaid bills. The API returns structured records; your workflow maps fields and creates a review report. Begin with read-only access. When you later create records, plan for duplicates, retries, audit evidence and approval. A connection that works once is not yet a dependable process.
11. Hosting and cloud servers: know what keeps the workflow running
Hosting means running your application or workflow somewhere that remains available. Managed services operate much of that environment for you. A virtual server gives you more responsibility for deployment, updates, access, backups and recovery.
Before moving a finance workflow online, ask who owns the account, where data is processed, what it costs, who receives failure alerts and how the system is restored. Your first learning project can stay small and supervised. You do not need to rent a server simply to learn what a trigger does.
Product logos identify the tools discussed and belong to their respective owners. Their use does not imply a partnership or endorsement.
The AI vocabulary worth understanding
- Context
- The instructions and information available to the model for a task. Give it the period, business meaning and approved source, not just “analyse this”.
- Tokens
- Units a model uses to process text and other inputs. They affect capacity and often API cost; they are not the same as words.
- API
- A structured interface between systems. An API key or OAuth grant determines what a connection can access.
- RAG
- Retrieval-augmented generation: retrieve relevant material, then use it to help generate an answer. Useful for a policy assistant, but retrieval quality, access rules and source dates still matter.
- Webhook
- An event notification sent from one system to another, such as “invoice approved”.
- JSON
- A common structured format for fields and values, useful when mapping invoice data between applications.
- Hallucination
- A plausible-looking model output that is incorrect or unsupported. Ask for evidence and verify material claims.
- Human approval
- A defined person checks and authorises an action before it affects the books, a payment or an external message.
A manageable four-week starting plan
- Week 1: clean a synthetic expense dataset and explain five checks you would perform.
- Week 2: use one AI assistant to analyse it, then reconcile and challenge the output.
- Week 3: connect one harmless input to a tracker with Make, Zapier or n8n.
- Week 4: document permissions, exceptions and what a reviewer must approve. Demonstrate the result in three minutes.
Technical confidence is becoming part of professional confidence. You still need accounting judgement, clear communication and ethics. The opportunity is to combine them with tools that make your work more useful.
Know someone who keeps asking where to start? Share this guide with your finance team or study group using the buttons below.
Create share images for Instagram & WhatsApp Status
Choose a feed card or a full-height story/status image. Download it, then post it from your app. Add the article link as a story link where supported, or paste it into your caption or status text.