K.
Practice — Where AI is already at work
kulu.co.za

Whatitactuallylookslike

Real engagements with property, marketing, accounting and HR firms across South Africa. Names are changed where the client asked. The numbers are exactly as they came out.

Engagements
37.
Industries
4.
Avg. days to live
12.
Still in production
100%.

Four firms, four shifts.

02 — Stories
Property — managing agent
Sandton

Tenant queries answered before the second cup of coffee.

Before

The team was processing 80–100 tenant queries every Monday. Most were the same five questions about levies, lease clauses and geyser repairs. Replies were taking two working days.

After

A drafting assistant trained on the firm's lease library and house style. Tenants get a draft answer in minutes; agents review, edit if needed, and send.

Result

Median response time down from 48 hours to 4. Same headcount.

Marketing — agency
Cape Town

Briefs to first drafts in a single afternoon.

Before

Senior strategists were spending half their week writing first-pass copy and decks for client work that would then be heavily rewritten.

After

An in-house drafting workflow trained on the agency's voice and tone guide. Strategists start from a 70% draft, edit to taste, ship to client review the same day.

Result

Strategist hours-on-drafting down by 60%. Client turnarounds halved.

Accounting & finance — mid-tier practice
Pretoria

Reconciliations that don't eat the weekend.

Before

Two staff doing reconciliations by hand each weekend, perpetually behind on month-end. Partner reviews bottlenecked everything else.

After

A Claude-based workflow drafts reconciliations directly from the bank exports, flags exceptions, and produces the working paper. Juniors review and sign-off.

Result

Month-end closing 4 working days earlier. Weekends returned.

HR — recruitment agency
Johannesburg

Long-lists triaged before the consultant opens the inbox.

Before

Consultants drowning in CVs for high-volume roles. The first read was eating 6–8 hours of partner-level time per role.

After

A structured screening assistant reads applications against the brief, scores against role-critical criteria, and surfaces a ranked long-list with notes.

Result

Time-to-shortlist down from a week to 36 hours. Placement rate held.

03 — A standard

Tested before it ships.

Every workflow we build is run against your firm's own real examples before it goes live. We measure what improved, what didn't, and we record where the human still has to sit.

When something doesn't work, we say so. When it does, the documentation is yours, the model context is yours, and the relationship continues.

04 — Yours, next

Would a story like one of these change something at your firm?

Tell us where the work is slow. We'll tell you, plainly, whether AI is the right answer.

Get in touch →