Key takeaways
- I use different AI tools for different jobs — drafting and reasoning, quick lookups, and occasional visual work — rather than one tool for everything.
- AI is useful for drafting variance narratives, explaining SAP configuration areas and stress-testing logic; it is not used to sign off controls or replace review.
- Speed lookups from any AI tool are a starting point, not a source — always verified against the actual standard, law or system documentation.
- The biggest time savings come from small, repeated tasks, not from one dramatic use case.
My "Top Ten AI Tools" list on the homepage — ChatGPT, Claude, Gemini, Microsoft Copilot, Perplexity, Midjourney, DeepSeek, Grok, Meta AI and Mistral AI — gets asked about more than almost anything else on this site. Most "top AI tools" lists stop at the name and the link. This post is the part that is actually useful: how a working finance manager uses these day to day, and, just as importantly, where I deliberately do not use them.
Drafting and reasoning: ChatGPT and Claude
These are the two tools I keep open most during a working week. Typical uses: turning a rough set of bullet points into a clear variance-analysis narrative for a management pack; drafting the first version of a policy or SOP section, which I then edit heavily against our actual process; talking through the logic of a reconciliation or a control design before I commit to it, the way I might think out loud with a colleague; and getting a plain-language explanation of an unfamiliar SAP FICO configuration area before I go into the actual configuration screens. The output is always a draft. Anything that touches a real number, a real control or a real filing gets reviewed and corrected by me before it goes anywhere.
Embedded in daily tools: Gemini and Microsoft Copilot
Because so much of finance work still lives in spreadsheets, documents and email, having Gemini and Copilot available inside the tools I already use — suggesting a formula, summarizing a long email thread, drafting meeting notes — saves more cumulative time than any single "big" AI task. The value here is friction removal, not new capability.
Fast lookups: Perplexity
When I need a quick starting point on a regulation, an accounting treatment question or an unfamiliar term, Perplexity's search-style answers with sources are a faster way to orient myself than a plain web search. The important discipline is what I say in the Editorial Policy for this whole site: an AI-generated summary is a starting point for research, never the final word for a regulated or tax-sensitive decision. Anything that will actually be used in a return, a statutory filing or an audit response gets checked against the primary source — the law, the standard, or a qualified adviser.
Occasional visual work: Midjourney
Not a finance tool at all — I use Midjourney occasionally for visual concepts in training decks or internal communication materials, where a simple illustrative image helps a slide land better than another bullet list.
Comparing alternatives: DeepSeek, Grok, Meta AI and Mistral AI
I keep these on the list because I genuinely use them, mostly to compare how differently a model answers the same finance or ERP question, and occasionally for cost or speed reasons on lower-stakes tasks. None of them get used for anything where the answer needs to be authoritative rather than a helpful second opinion.
A simple framework for finance teams considering AI tools
- Classify the task first: is this drafting/explaining (AI-friendly), or deciding/approving (stays with a qualified person)?
- Never let AI touch the control: use it to prepare the analysis, never to be the reviewer or approver of record.
- Verify anything regulated: tax, VAT, statutory and legal questions get checked against the primary source before use.
- Measure time saved on repeated tasks, not on one impressive demo — that is where the real productivity gain shows up over a quarter.
Where this is going
AI tools have not changed what finance is for — turning data into decisions that can be trusted. They have changed how much of the mechanical work between "raw data" and "a defensible answer" I have to do by hand. That is a genuine gain, as long as the judgment, the review and the accountability stay firmly with a person.
Written by Liakat Hossain
Finance and accounts professional with experience in FP&A, SAP FICO, Power BI, internal control, reporting, audit, tax, VAT and export-oriented business operations.
This is an original blog post written by Liakat Hossain, drawing on his own professional and personal experience. It is a personal reflection, not regulatory, tax, legal or investment advice; verify any technical or professional point against current standards before relying on it.