Automating the monthly B-BBEE report without losing control

Most transformation managers I know have a version of the same month. The first week goes to collecting: the payroll export from HR, the spend report from finance, the training register from the skills development facilitator, the renewed supplier certificates from procurement. The second week goes to reconciling them. By the time the committee report is drafted, the figures are three weeks old, and the person who built them has had little time to ask what they mean.

That month can be automated. Most of the work is mechanical, repetitive and governed by rules. But a monthly B-BBEE report is a set of figures that management will act on, and that will eventually be tested at verification. Automating it carelessly swaps a slow, visible process for a fast, invisible one. Given that choice, I would keep the slow one.

You do not have to choose. Here is how I think about automating the report while keeping control of it.

Three kinds of work

When we map a client's monthly reporting, the work falls into three groups.

Collection and calculation. Pulling data from source systems, reading new evidence, applying the scorecard rules for the Codes or sector code that applies to you, and producing the figures. This should be automated as fully as possible. It is where errors creep in when people do it by hand, and where a well-tested system is most reliable.

Drafting. Turning the figures into commentary: what changed since last month, why it changed and what needs attention. AI writes a good first draft of this, provided it works only from confirmed figures and shows where each statement comes from.

Judgement and approval. Deciding which risks matter, what to recommend and whether the report is right. This stays with people, every month.

Most of the time saved comes from the first group. Most of the risk sits in the second. The third is the reason the report exists at all.

Five controls I would not give up

1. Every figure traces to its source

Each number in the report should link to the records that produced it, and those records to their source documents. If a director asks why preferential procurement fell, the answer should be a click away: these suppliers, these invoices, this certificate that lapsed in August.

A report you cannot trace is a report you have to trust, and trust is not a control.

2. Reconcile to the books

Automated figures should be checked against independent totals every month: measured procurement spend against the general ledger, headcount against payroll, skills spend against the training budget. Differences are shown, not smoothed over. A system that never reports a reconciliation difference is usually hiding one.

3. Exceptions go to a named person

The system should raise anything it cannot resolve on its own: a supplier without a valid certificate, an invoice it cannot match, a learner without a signed agreement, a figure that moved further than expected. These form a short queue, and a named person clears each item with a recorded decision. That is where your transformation champion's time is best spent.

4. AI drafts, people edit, one person approves

The commentary remains a draft until a person has read it against the figures, edited it and approved it. The approval is recorded with a name and a date. AI-written text should never reach a committee unreviewed, however well it reads. Fluent writing is precisely what makes an unchecked error dangerous.

5. Each month is locked once approved

Once a report is approved, its figures are locked. Later corrections are made as new entries with a reason, not by quietly changing last month's numbers. That way the report the committee saw in July can always be reproduced, and the path from July's projection to the final verified level is visible to anyone who asks.

What changes for the team

When this works, the month looks different. Data flows in continuously instead of being chased in the first week. The figures update as evidence arrives. The champion spends a day on exceptions and an hour reviewing the draft, rather than a fortnight on assembly. The committee receives a report that is current, consistent with the one before it and traceable to its sources.

The bigger gain is that a report produced the same way every month becomes comparable. Trends show up. A skills development figure falling behind its target is visible in the first half of the year, while there is still time and budget to respond.

Where to be careful

Automation repeats whatever rules you give it. If the scorecard logic is wrong, or applies the generic Codes where a sector code governs you, an automated report will repeat the mistake every month with complete confidence. Before switching anything on, I would have the rules checked by someone who knows the Codes well and tested against a period for which you already know the verified result.

Personal information also needs care. A monthly report rarely needs to name anyone. Build it from totals, keep individual records behind role-based access, and keep the system where your other confidential records are kept.

If your team loses most of every month to building the B-BBEE report, we would be glad to map the process with you and show where automation would help and where it should stop. Write to me and the team at sales@bsvision.co.za.