AI strategy and readiness
A readiness assessment, a prioritised roadmap and a business case per use case.
AI consulting and implementation for South African organisations
Arke Technologies (Pty) Ltd is a Johannesburg AI consulting and implementation company. We help mid-sized and large organisations, public-sector bodies and schools in South Africa choose the AI use cases worth funding, then build, integrate and govern them, with POPIA controls, audit logs and people in charge of every decision that matters.
What AI consulting is, and what an AI consultant does
AI consulting is the advice and engineering that turn artificial intelligence from a tool a few people try into a system an organisation can rely on, defend to its auditors and afford to run. A useful AI consultation starts with the work, not the technology: which decisions and documents take the most skilled time, and what a wrong answer would cost.
AI consultants then do four jobs. They prioritise use cases by value and risk, design the AI solutions architecture (models, data sources, integrations and controls), build and integrate the system, and set up the monitoring that keeps it accurate after launch. Our AI consulting services cover all four, so the team that writes your AI strategy is the team accountable when it runs.
Buyers meet several kinds of supplier. An AI agency, or artificial intelligence agency, tends to focus on marketing content and chat experiences; management consulting firms often treat AI as one workstream in a wider strategy programme; an AI tech company usually sells one product. Arke Technologies advises, builds and then supports, with software development, systems integration, cybersecurity and managed IT in the same company, so an AI system lands on infrastructure we already know how to secure and run.
Our AI consulting services
Each can be bought on its own or as one programme with one accountable team.
Microsoft 365 Copilot
Licensing, permission clean-up, sensitivity labels, a measured pilot and adoption.
Generative AI and LLM integration
Large language models connected to your documents and systems, with guardrails.
AI agents
Software that plans and completes multi-step tasks inside approval limits you set.
AI automation
Workflows where AI reads, classifies and drafts, and people handle the exceptions.
Chatbots and conversational AI
Assistants for customers, citizens, learners and staff that hand over to a person.
Document intelligence
Extraction and classification for invoices, contracts, forms and applications.
Private AI
Open-weight models on GPU servers you control, on your site or in South African colocation.
AI governance
Policies, risk registers, approval points and audit trails a board can oversee.
AI intelligent agents: definitions and examples
An intelligent agent in artificial intelligence is software that perceives its environment, decides what to do, acts towards a goal and checks the result. The study of artificial intelligence and intelligent agents is decades old; large language models now let an AI intelligent agent follow plain-language instructions, call business systems and explain what it did.
Intelligent agents in artificial intelligence, with examples organisations can run today:
- A service-desk agent that reads a ticket, gathers the device history and proposes a fix for a technician to approve.
- A sourcing agent that gathers dated price evidence and drafts a quote for a person to check, the pattern our AI-assisted tools on the Server Hub platform follow.
- A compliance agent that watches filing deadlines across a client portfolio and drafts reminders.
- A procurement agent that compares bids against a specification and flags gaps for the evaluation committee.
AI intelligent agents earn their keep when their permissions are narrow and every action is logged. See AI agents for business and our explainer on AI agents vs generative AI.
The models and platforms we build with
Generative AI is the family of models that produce text, images, code or audio: mostly large language models today, with image models and older generative adversarial networks (GANs) used for specialist tasks such as synthetic test data. Classical machine learning still does much of the useful forecasting and classification work, at a fraction of the compute.
We build with leading models from Anthropic (Claude) and Google (Gemini, through Vertex AI), and with open-weight models, choosing the model for each job. On Google Cloud, ML work runs through Vertex AI and BigQuery; we also build on Microsoft Azure and AWS where your estate already lives. Retrieval-augmented generation (RAG) grounds answers in your own documents, and guardrails, logging and cost controls ship with every system. Where data must stay on hardware you own, we run open-weight models on a private AI platform.
Adoption and training
Adoption decides the return on any AI investment. Business schools offer AI for business masters programmes, certification courses and a growing shelf of books, but most staff need something narrower: short, role-based training on the tools they actually have, clear rules on what may go into a prompt, and a simple way to report a wrong answer.
Our AI consulting training is built around your use cases: practice on your own documents, the limits of each tool and how to check an output before it leaves the building. We measure adoption through usage reports and short surveys, report it in the monthly service review, and adjust the training where uptake is low.
Governance, POPIA and responsible AI
Our Emergent Intelligence (EI) framework is the set of design principles we apply so that AI strengthens human judgement rather than replacing it. It has five principles:
Agency preservation. AI recommends; people decide, above all where a decision affects someone's rights, money or health.- Transparency. Every output can be explained: what data informed it and why.
- Governance and auditability. Every AI action is logged and reviewable.
- Progressive enhancement. Start with low-risk uses and expand as confidence builds.
- Ethical boundaries. Clear limits on what the AI may do and what data it may see.
Data handling follows the same discipline: data minimisation, encryption in transit and at rest, access controls and audit logging. We never train models on client data without explicit consent.
The law points the same way. POPIA requires appropriate security for personal information, restricts decisions based solely on automated processing that significantly affect a person (section 71), and limits transfers of personal information abroad (section 72). King IV Principle 12 makes the governing body responsible for technology and information governance, which reaches AI systems as it reaches any other technology, and public bodies add PFMA or MFMA disciplines on procurement, records and audit. Our security controls are aligned with ISO/IEC 27001 and the NIST Cybersecurity Framework. Read our POPIA compliance checklist, King IV and technology governance, our guide to South Africa's national AI policy and how we disclose our own use of AI.
How an AI engagement is governed and reported
- A named lead for your engagement and a steering group that meets on an agreed cadence.
- Change control for prompts, model versions and data sources, so no model update goes live unreviewed.
- A monthly service review and report once a system is in production: accuracy on sampled outputs, usage, cost against budget, incidents and requests.
- Incidents prioritised P1 (critical outage) to P4 (low-impact request), with response and resolution targets, escalation paths and exclusions agreed in your SLA.
- An incident route for AI errors: a harmful or wrong output is logged, traced to its prompt, data or model version, and the fix recorded.
- Testing of the AI's access to your systems by our cybersecurity team, including penetration testing where a system is exposed to the internet.
Procurement fit and continuity
- Level 1 B-BBEE contributor (sworn affidavit available on request), which gives buyers 135% procurement recognition.
- Supplier documents. Arke Technologies (Pty) Ltd has been a registered company since 2017 (registration 2017/386810/07). Company documents, bank confirmation letter, CSD supplier number and SARS tax compliance status are available on request, and we complete standard bidding documents (SBD forms) when a bid requires them. See procurement and supplier compliance and our company facts.
- How to buy. A proposal for the readiness assessment, then an SLA or project agreement against your purchase order, or our response to your tender.
- You own what we build. Your data, prompts, configurations, code and documentation stay yours. Systems are documented so another team could run them, and on exit we hand over everything, credentials included.
- Model independence. We design so the model can be changed without rebuilding the application, which protects you when prices or providers change.
- One accountable team. Strategy, build, security and support sit under one contract with Arke, so there is no gap between the people who designed a system and the people who answer for it in production.