Ask most owners how last month went and you get an answer about the bank balance. It went up, so it was fine. Or it went down and there was a big supplier bill, so that explains it.
Sometimes that answer is right. Often it hides something. Revenue was up because one large customer paid two invoices at once, while the number of new customers fell for the third month running. Margins on one service quietly slipped because a supplier raised prices in the spring and nobody repriced. The bank balance can't tell you any of that. It is the sum of everything, and the sum is where the detail goes to disappear.
Pulling that detail back out is what a business analyst is for. This post covers what a business analyst does, what a business analyst job description usually asks for, how the role differs from a data analyst, and which parts of the job an AI can run on a schedule. We make Sanaf, an AI employee that can take on the recurring half of this work, so we have an interest in that last section. The product shows up near the end, and everything before it stands on its own.
The short answer
A business analyst works out how a business is actually doing and what it should change. They gather numbers and information from across the business, find what moved and why, spot problems and opportunities, and write it up so the people who decide can decide. In larger companies the title also covers analysts who gather requirements for new systems and processes. In most businesses the job is closer to this: someone who reads the numbers every week, explains them in plain words, and says what looks wrong.
What does a business analyst do all week?
The title covers a lot of ground, so it helps to split the work into the things that actually fill a week.
1. Gathering the numbers
Before any analysis there is collection. Sales sit in the CRM. Revenue and costs sit in the accounting software. Website visits sit in analytics. Job or order records sit in a spreadsheet or an operations tool. A business analyst knows where each number lives, pulls them on a regular rhythm, and lines them up so that "March" means the same thing in every source.
This sounds like admin, and it is. It is also where most reporting dies, because a monthly pull from four systems is exactly the kind of task that slips the week something urgent comes in.
2. Watching the handful of numbers that matter
Every business has a short list of key performance indicators, or KPIs, that tell you whether things are healthy. For a service business it might be new enquiries, quote-to-win rate, average job value, gross margin and cash collected. For an online store it might be orders, average order value, return rate and repeat customers. A good analyst keeps that list short and watches it every week, so a change shows up while there is still time to act on it.
3. Telling a trend from a one-off
This is the skill that separates analysis from a dashboard. A number that has crept up for six weeks is a trend. A number that jumped once last Tuesday is an anomaly. They call for different responses. A trend asks "what changed in how we work?" An anomaly asks "what happened that day, and is the data even right?" An analyst names which one they are looking at, because treating a blip like a trend is how businesses overreact, and treating a trend like a blip is how they miss the slow leak.
4. Finding the reason
Once something has moved, the analyst digs into why. Revenue is down: is it fewer customers, smaller orders, or one big account that left? Margin is down: is it prices, costs, or a shift toward a cheaper service line? This is usually done by breaking the number apart by customer, product, location, salesperson or month until the change sits in one place.
5. Writing it up
The output of all this is a report, and a useful one is short. It's a page of writing that says what happened, why, how sure the analyst is, and what they would do about it. A business analyst who finds the right answer and can't explain it in three paragraphs has done half the job.
6. Recommending, and leaving the decision with the owner
A good analyst ends with a recommendation: reprice this service, chase these overdue accounts, look at why enquiries from the website dropped. Then they stop. The decision belongs to whoever runs that part of the business, because the analysis is an argument and someone has to be persuaded by it.
7. The bigger-company version
In larger organisations, "business analyst" often means something more specific: the person who sits between a business team and the IT team, gathers requirements for a new system, maps current processes, writes user stories, and checks the delivered system does what was asked. If you are reading a job ad from a bank or a software company, this is often the role. If you run a business and want someone to tell you how it's going, it usually isn't.
Business analyst responsibilities, as a job description lists them
Read a handful of business analyst job descriptions and the same business analyst responsibilities keep coming up:
- Collect and combine data from the company's systems.
- Define and track KPIs with the leadership team.
- Analyse revenue, sales performance, costs and margins.
- Identify trends, risks and opportunities.
- Investigate unusual results and explain them.
- Produce regular reports, weekly or monthly, for managers and owners.
- Recommend changes to pricing, process or focus.
- Map current processes and document requirements for new ones.
- Present findings clearly to people without a finance background.
The business analyst skills listed underneath are usually some mix of spreadsheet fluency, comfort with reporting tools, basic statistics, clear writing, curiosity, and the nerve to tell a manager their favourite project isn't working.
As with most roles, the list splits into two kinds of work. Some of it is judgment: which KPIs matter for this business, what a change means, what to recommend, how to say something unwelcome. Some of it is repetition: pull the same numbers, build the same table, check the same thresholds, write the same weekly summary. That split decides which parts an AI can do well.
Business analysis techniques worth knowing
You don't need a certificate to use the common business analysis techniques. Most of them are structured ways of asking good questions.
- Variance analysis. Compare actual results with a budget, a forecast or the same period last year, then explain the gap. This is the backbone of a monthly report.
- Root cause analysis. Keep asking "why?" about a problem until you reach something you can change. The "five whys" is the simplest version.
- Segmentation. Break a total down by customer, product, region or channel. Most surprises live inside one segment.
- Gap analysis. Describe where the business is, where it wants to be, and what stands between the two.
- SWOT. Strengths, weaknesses, opportunities and threats. Useful once a year, less useful as a weekly habit.
- Process mapping. Draw how a piece of work actually moves through the business, step by step, to find the delays and handoffs where it stalls.
Business analyst vs. data analyst
People mix these two up constantly, and job ads don't help. Here is the practical difference.
| Business analyst | Data analyst | |
|---|---|---|
| Starts from | A business question: why did margin drop? | A dataset: what does this data show? |
| Main output | A written finding and a recommendation | Queries, models, charts and datasets |
| Typical tools | Spreadsheets, accounting and CRM reports, slides | SQL, Python or R, BI tools |
| Talks most to | Owners, managers, operations | Other analysts, engineers, data teams |
| Measured on | Better decisions | Accurate, usable data |
In a large company the two work side by side: the data analyst builds a reliable dataset, the business analyst turns it into a decision. In most businesses there is neither, and the real question is whether anyone reads the numbers at all.
What a good business report looks like
A useful weekly or monthly report is short and follows the same shape every time, so the reader learns where to look. Here is a business report example in that shape. The business and the numbers are made up for this post.
Weekly numbers, week ending Friday
Headline. Revenue is in line with the last four weeks. New enquiries are down for the third week running, which is the thing to look at.
What moved. Enquiries from the website form fell from roughly 30 a week to under 20. Phone and referral enquiries are flat. The drop started the week the contact page was changed.
What looks odd. One invoice was entered twice in the accounting software on Tuesday. Revenue above already excludes the duplicate.
How sure I am. Confident on revenue and enquiries. Less sure on margin, because two supplier bills for the period haven't been entered yet.
What I'd do. Check the contact form on mobile. Ask the bookkeeper to remove the duplicate invoice.
Five parts: the headline, what moved, what looks odd, how sure, what to do. The fourth part is the one most reports leave out, and it is the one that stops a business acting on a half-finished number.
How much does a business analyst cost?
There is no honest single number to put here, so we won't. Pay depends on seniority, location, industry and whether the role is full time, part time or a contractor on a monthly retainer. Look at current pay for the role in your area on a job board or a salary survey before you budget.
The costs people forget are easier to name:
- Access. An analyst needs read access to accounting, CRM and analytics. Setting that up and keeping it tidy takes someone's time.
- Your time. The report is only worth something if somebody reads it and acts. Budget an hour a week for that, whoever writes it.
- Gaps. When the analyst is on holiday, the report usually stops. Bad months tend not to wait.
For many businesses the honest comparison isn't with a hired analyst at all. It's with the owner, rebuilding the same spreadsheet on a Sunday night, or more often not rebuilding it.
Where AI for business analysts actually helps
Search for AI for business analysts, or AI tools for business analysts, and you'll find three different kinds of product mixed together:
- Chat over a spreadsheet. You upload a file and ask questions about it. Handy for a one-off question. It knows nothing about your business beyond the file, and it only runs when you remember to ask.
- AI features inside BI and dashboard tools. Natural language queries, suggested charts, automatic summaries of a dashboard. Useful if you already have the tool and someone who opens it.
- An AI business analyst. Something that takes on the recurring job: pulls the numbers from the systems you connect, watches the KPIs you named, writes the weekly report, flags anomalies and recommends what to look at. This is the one that changes the "nobody looks" problem, because the report turns up whether or not anyone remembered.
The third kind fits the repetitive half of the job well. It can:
- Pull the same numbers from accounting, CRM, analytics and spreadsheets on a schedule.
- Track the KPIs you chose, every week, without being asked.
- Separate trends from one-off anomalies and say which is which.
- Break a change down by customer, product or month to find where it sits.
- Write the weekly or monthly report in plain words.
- Recommend what to look at first.
The judgment half should stay with people, and a well-behaved AI makes that easy:
- It says when it is unsure. A partial month, two sources that disagree, a category renamed halfway through the quarter: all said out loud. A confident wrong number is worse than no number, because you act on it.
- It recommends and stops there. It does not reprice, cancel, spend or announce anything because of an analysis.
- It shows its working. You can see which numbers it used, from where, and when they were pulled.
- It does not pick a winner quietly. When your accounting software and your CRM disagree, that's a question about how your business counts things, and the answer is yours.
Will AI replace business analysts?
For businesses that already have a good one, no. The valuable part of a business analyst is the judgment: knowing which question matters this month, reading context the numbers don't hold, and persuading a manager to change course. An AI that takes the weekly pull-and-write off their plate gives them more time for that. For businesses that have never had an analyst, the realistic change is different. The numbers finally get read every week.
How to start reading your numbers every week
If your reporting today is the bank balance and a feeling, here is a simple way to start, with or without help:
- Pick five numbers. One for demand (enquiries or visits), one for sales (won deals or orders), one for value (average job or order), one for profit (gross margin), one for cash (money collected). Five is enough.
- Write down where each one lives. Which system, which report, who has access.
- Choose the comparison. Last week, the four-week average, or the same week last year. Pick one and stick with it.
- Set a threshold for "look at this". For example, any number more than 15 per cent off its four-week average.
- Use the report shape above. Headline, what moved, what looks odd, how sure, what to do.
- Read it at the same time each week. Monday morning is common. Make one decision from it, even a small one.
Whoever ends up writing the report, a person, a contractor or an AI, will do a much better job with steps 1 to 4 written down.
Where Sanaf fits
Sanaf is an AI employee. It works in your team's Slack or Microsoft Teams and inside the apps you connect. Business analysis is one of the jobs it can take on, and the AI business analyst page sets out exactly what that job covers. The AI employee roster shows the others.
In practice it does the recurring half described above. It monitors the KPIs you name, analyses revenue and sales performance, identifies trends, detects anomalies, writes the business report and recommends actions. It reads from QuickBooks, HubSpot, Google Analytics and Google Sheets, and only from the ones you connect. You can ask in a sentence, like "send me the weekly numbers every Monday at 8, and tell me if enquiries drop more than 15 per cent", and the report arrives in the same channel.
The limits are the ones listed earlier. It reports what the numbers say and where it is unsure, and it never dresses up a shaky number as a certainty. Every recommendation is a person's to accept or ignore. Anything it can't undo asks first, and each job leaves a work log the team can read. If the books themselves need keeping, that is a separate job, covered on the AI bookkeeper page. If you want the sales pipeline cleaned before it gets analysed, the AI sales operations post covers that half. And if business analysis is one job among several, the AI chief of staff post explains how Sanaf hands work to agents that report to it.
On price, you start with $120 of free credit and no card, and that free credit lasts 30 days. After that you buy credits from $20 when you need them, and credit you buy never expires. You pay for the work, not per seat. The pricing page has the details, and hiring an AI employee shows what the first day looks like.
A good first test: connect your accounting software and one spreadsheet, name five KPIs, and ask for last month's report. Then check it against what you already know about that month.
Questions people ask about business analysts
What does a business analyst do?
A business analyst works out how a business is performing and what it should change. They gather numbers from the business's systems, track key performance indicators, find what moved and why, separate trends from one-off anomalies, write reports and recommend actions for managers to decide on.
What are a business analyst's main responsibilities?
The main business analyst responsibilities are collecting and combining data, defining and tracking KPIs, analysing revenue, sales and margins, identifying trends and risks, investigating unusual results, producing regular reports and recommending changes. In larger companies the role often adds process mapping and gathering requirements for new systems.
What should a business analyst job description include?
A good business analyst job description names the systems the analyst will read from, the KPIs they will own, how often they report and to whom, whether they map processes or gather system requirements, and who makes the final decision on their recommendations.
What skills does a business analyst need?
The core business analyst skills are spreadsheet and reporting-tool fluency, basic statistics, clear writing, curiosity about why a number moved, and the confidence to deliver an unwelcome finding. Knowledge of the industry helps more than any single tool.
What is the difference between a business analyst and a data analyst?
A business analyst starts from a business question and ends with a recommendation for managers. A data analyst starts from data and ends with accurate queries, models and datasets. In large companies they work together. In most businesses the business analyst role matters first.
What is an AI business analyst?
An AI business analyst is an AI that takes on the recurring part of business analysis: pulling numbers from connected systems, tracking chosen KPIs, spotting trends and anomalies, writing reports and recommending actions. A good one says when it's unsure and leaves every decision to a person. Sanaf works this way.
Will AI replace business analysts?
AI is unlikely to replace good business analysts, because the valuable part of the role is judgment and persuasion. It can take over the weekly collection and reporting work, which gives analysts more time for that judgment and gives businesses without an analyst a report every week.
What should a weekly business report include?
A useful weekly business report has five parts: a one-line headline, what moved, what looks odd, how sure the writer is about each number, and what they recommend doing. Keeping the same shape every week lets the reader spot changes quickly.