Skip to content
digital army.ai digital army.ai
Start a scoping conversation
Autonomous AI for regulated enterprise

The firm a corporate brings in when an autonomous AI decision has to hold up.

A board, an auditor, a regulator. Building an agent has become cheap; the judgment to know what should be built — and the standing to defend it when the hard question comes — has not. That judgment is what we sell.

Redefining work with digital employees

  1. R2.55m
    Recovered every month
    By a single collections agent, for a leading South African mobile network operator.
  2. In production
    Live across three sectors
    Autonomous systems running in financial services, telecommunications, and recruitment.
  3. Level 1
    B-BBEE contributor
    A South African firm, building for regulated enterprises across the country.
The shift

The build is no longer the hard part.

A configured platform now does in weeks what a skilled team once took a quarter to build. The ability to build has become common — so it is no longer where the value sits.

The risk, though, outlives the build. An ungoverned agent carries the full weight of production risk, and ungoverned agents spread through a business faster than anyone can account for them —

until the morning one fails at scale, breaches a mandate, or makes a decision someone has to explain to a regulator.

We are engaged for the scarce part: knowing what should be built, and standing behind it once it is running.

The record — autonomous systems in production

Built to a number, and held to it.

Not a product count. Systems built by DigitalArmy, running on real client data, each aimed at a named business result and measured against it. Status is stated plainly.

  1. Built by DigitalArmy and since handed to the operator’s in-house team with the audit trail and monitoring intact. A failed debit order is the start of a quiet revenue leak; the agent reaches the customer within hours, reinstates service the moment payment confirms, and escalates only the complex cases to a human specialist.

    Recovered / monthR2.55m
    Response rate22%
    Recovered / customerR87.13
    StatusProduction · handed over

  2. Credit applications used to die in the gap between an interested customer and a complete document pack. The agent engages the applicant directly, asks for what is needed, validates what is sent, and returns a complete vetting pack — governed under POPIA from the first sub-phase.

    Referrals → sales55%
    Docs auto-collected90%
    StatusIn production

  3. Most digital sales journeys leak in the middle. The agent re-engages on the original channel and references the specific session — incremental revenue from customers who were about to be lost, with no discounting required.

    Conversion (responders)21%
    Overall recovery7%
    StatusIn production

  4. Recruitment funnels lose more candidates to admin friction than to unsuitability. The agent captures details, screens against the role, collects documents, and books the next step — all inside one WhatsApp conversation a candidate can finish on their phone.

    Application completion40%
    StatusIn production

Beyond these, the firm is engaged from strategy through pilot to build across telecommunications distribution, fundraising, and regulated healthcare. Full records available on request.

See the evidence →
How we work

We meet an organisation where it stands, and build toward systems it can trust and own.

Before any process is automated, it is improved under conventional discipline — automating a broken process produces broken automation at scale. The first delivery is always a working slice on real data, and governance is engineered into the architecture from the first sub-phase.

01 Board

Governance & oversight

The AI governance framework, ISO/IEC 42001 readiness, and a Fractional Chief AI Officer where senior governance presence is needed without a permanent appointment.

02 Executive

Strategy & prioritisation

The sourcing posture, and the high-value workflows automated first.

03 Management

Delivery

Fix-First discipline, the build, and a reporting cadence the board can act on.

04 Workforce

Literacy & access

Practical coaching in sanctioned tools, and safe-use guidance that addresses unmanaged tooling.

05 The core

Autonomous execution

Digital employees that run real work on real data, and the Decision Record the client owns once they are live.

What we do, in full →
Accountability, engineered in

Every autonomous decision leaves a record the client owns.

For sensitive work — credit, claims, candidate evaluation — the agent flags borderline and high-risk cases for a person to decide. Every decision produces the Decision Record: a permanent, signed, timestamped account of the inputs the agent read, the boundary of permitted action that applied, and the action it took. The client owns it, can inspect and export it, and can take it in-house at any time. Where an engagement warrants it, the work is committed to ISO/IEC 42001 — formal proof of governance for a regulator, an insurer, an investor, or a major customer.

Decision Record Collections Agent · one decision, on the record Client-owned
1 Inputs readAccount status, payment history, mandate terms, and the customer’s contact channel.
2 Permitted boundaryOffer pre-approved repayment options only. No rate change, no write-off.
3 Action takenIssued repayment options on WhatsApp; reinstated service the moment payment confirmed.
4 Logged & signedAgent v3.2 · 2026-02-04 07:14 SAST. Escalation: none — resolved within mandate.
Signed · timestamped · exportable in an open format
The senior core

The people who scope it have sat in the chair the client sits in now.

The firm is built around a bound senior core of operators who have run enterprise technology at the scale of a major telecommunications group and a major insurer, and answered to a board and a regulator for it. That is what a corporate engages us for: a senior peer in the room who reduces the executive's personal exposure. Delivery scales through a network of specialist developers under that senior architecture.

Meet the team →
  • 01
    Telecommunications at scaleEnterprise technology run at the scale of a major South African telecommunications group.
  • 02
    Regulated insuranceCIO-level leadership across a regulated UK and South African insurer.
  • 03
    Board & regulator accountabilityOperators who have answered to a board and a regulator for the systems they ran.
  • 04
    Specialist delivery networkDevelopers delivering under the senior architecture, so quality and response time hold as engagements grow.
Capability

Platform-agnostic by design, so the client keeps a stack it can exit.

Tooling is chosen per task, matched to the smallest model and the simplest pattern that meets the metric. We hold the partnerships and deploy them; we are not tied to any one vendor.

  • Stubber The South African AI-employees platform that runs several of the firm’s production agents.
  • Microsoft Copilot Studio, Azure AI Foundry, and the Power Platform, delivered with the firm’s Microsoft delivery partner.
  • Matogen Data, low-code, and applied-AI capability where the data substrate is the load-bearing part.
  • XOPS Autonomous IT operations.
  • Nuacem Conversational and voice AI, deployed across South African businesses.
How we engage

Where the metric can be underwritten, we price on the result.

Engagements run in stages — each one produces the evidence on which the next is committed, so the programme is funded against results rather than forecasts.

  1. 01

    Qualification & scoping

    A working session that maps the process and baselines the opportunity.

  2. 02

    Working product

    A vertical slice running on real client data — evidence, not slideware.

  3. 03

    Production build

    The system, engineered to scale, with governance as an output.

  4. 04

    Annuity

    The system is operated and improved; priced on the result where the metric can be underwritten.

Common questions

What a senior buyer asks first.

Who is DigitalArmy for?
Executives in regulated enterprises — financial services, insurance, telecommunications, and healthcare — accountable for autonomous decisions that must hold up under a board, an auditor, and a regulator.
Why do most enterprise AI projects fail?
They automate a broken process, or run agents no one can account for. We fix the process first and engineer governance into the architecture from the first sub-phase, so the system holds up when it is questioned.
How do you prove an autonomous decision to a regulator?
Every autonomous decision produces the Decision Record — a signed, timestamped account of what the agent read, the boundary that applied, and the action it took. The client owns and can export it. Where it warrants, work is committed to ISO/IEC 42001.
What does an engagement cost?
Engagements run in stages — scoping, a working product on real data, a production build, then an annuity. Each stage funds the next against evidence. Where the client can underwrite the metric, the firm prices on the result.
Where is DigitalArmy based?
Johannesburg, Gauteng — working with regulated enterprises across South Africa.
Start here

Start with a scoping conversation.

A 90-minute working session, not a pitch. We map the process, baseline the opportunity, and define what success would have to look like — the inputs to a paid diagnostic and a working product, if it makes sense to proceed.

Office
Johannesburg, Gauteng
Hours
Mon – Fri · 08:00–17:00 SAST
Reply
A senior member of the team reads every enquiry, and responds within one business day.

Tell us the process you would like to examine and we will arrange a 90-minute working session — a senior member of the team responds within one business day.

Start a scoping conversation

Your enquiry routes privately to the firm. Submissions are held only for the duration of the conversation.