Sentient Software is a South African engineering firm that designs and deploys autonomous AI agents to take over repeatable operational work — campaigns, bookings, reporting, and data workflows.
These aren’t chatbots. They’re production systems with tools, guardrails, approvals, and failure handling — built to run without babysitting.
Based in South Africa. Working with teams worldwide.
How this works
Agents don’t “answer questions” — they execute workflows. They connect to your tools, follow rules, request approvals when needed, and keep going until the job is done.
Operates like a service, not a one-off interaction.
Calls APIs, updates records, triggers automations.
Tracks context, applies policies, follows playbooks.
Retries, escalates, and logs every step.
Where this applies
Deployed systems that execute work across your tools — with rules, approvals, and observable runs.
Turn a brief into live ads and ongoing optimisation — without babysitting.
Handle inbound requests end-to-end: qualify, schedule, confirm, and follow up.
Generate the weekly numbers, flag anomalies, and ship updates to the right people.
The simulation above shows the run experience. Real deployments connect to your systems and operate under your rules.
Built for ops, growth, and revenue teams who need workflows to run reliably — not just generate suggestions.
If the process is repeatable, structured, and currently annoying, it’s probably a good fit.
Fit checklist
Rule of thumb: if a human follows a checklist and clicks around 3–5 tools, an agent can run it reliably.
A production workflow: connected to your tools, constrained by your rules, observable from day one.
Common questions
Short answers. The Solutions page covers guardrails, integrations, timelines, and pricing in more depth.
No. Agents don’t answer questions — they execute workflows. They connect to your tools, follow rules, request approvals when needed, and keep going until the job is done.
It runs continuously like a service rather than a one-off interaction, uses tools and systems by calling APIs and updating records, maintains state and applies rules and playbooks, and handles failure modes through retries, escalation, and logging every step.
A repeatable workflow with clear inputs and outputs, multiple tools involved such as a CRM, calendar, ad platform, or reporting system, and rules that can be written down, such as SLAs, budgets, or policies. Exceptions can exist as long as most cases are standard. Rule of thumb: if a human follows a checklist and clicks around 3–5 tools, an agent can run it reliably.
One-off creative or open-ended strategy work, processes with no source-of-truth systems to operate on, unclear success criteria or constantly changing rules, and pure “write content” tasks with no execution loop.
We’ll pressure-test the workflow, identify where autonomy actually helps, and be direct if it’s not a good fit.
Or email: warren@sentientsoftware.co.za