How an engagement actually runs
Four steps: Discover, Design, Build, Automate. Each one has a defined output, so you always know what you are paying for and what arrives at the end of it.
Quick answer
An AiEngineer.in engagement runs in four steps. Discover audits your workflows, systems, and data access. Design produces a scoped roadmap with success metrics. Build delivers the agents, automations, and integrations against that scope. Automate takes the system live and keeps improving it with monitoring, reporting, and monthly iteration.
The four steps
Discover and Design decide what is worth building. Build and Automate deliver it and keep it working.
- 1
DISCOVER
AI audit of your workflows & data
- 2
DESIGN
Custom AI roadmap & automation plan
- 3
BUILD
Agents, chatbots & integrations delivered
- 4
AUTOMATE
24/7 AI running your operations & scaling
DiscoverUnderstand the work before proposing any technology
We start with how the work happens today — who touches it, which systems hold the data, and where the queue backs up. The free 30-minute AI Opportunity Audit is the entry point; if you proceed, Discover goes deeper into volumes, exceptions, and system access.
What we need from you
A few hours from one operational owner who knows how the work really happens, plus read access to reports or exports where relevant.
What happens
- Walkthrough of the target workflows with the people who run them
- Volume review: tickets, leads, messages, and documents per month
- System inventory and check on API or export access for each tool
- Exception mapping — the cases that break the happy path
- Baseline capture: current response times and hours spent per week
What you receive
- Written summary of each workflow as it actually runs today
- Shortlist of automation candidates ranked by volume and feasibility
- Honest note on anything we would not recommend automating yet
DesignA scoped plan with success metrics, before anyone writes code
Design turns the shortlist into a build plan: what the AI is allowed to read and write, where a person must approve, which integrations are required, and how success will be measured against the baseline recorded in Discover.
What we need from you
Review and sign-off on scope, tone, and escalation rules, plus a decision on which phase goes first.
What happens
- Automation design per workflow, including escalation and failure paths
- Data handling decisions: what the AI sees, stores, and never touches
- Integration plan per system, with fallbacks where no API exists
- Success metrics agreed against your recorded baseline
- Phasing so the first release is small enough to verify quickly
What you receive
- AI roadmap document with scope, design, and human checkpoints
- Integration plan naming each system and access method
- Fixed written quote and delivery schedule per phase
BuildShip one workflow at a time and prove it in your environment
Build delivers against the agreed scope: agents configured on your content, automations wired to your systems, and testing in a staging setup with your real cases. Nothing goes live until the escalation and approval paths have been exercised.
What we need from you
Access provisioning, a nominated tester, and sign-off after reviewing the staging results.
What happens
- Agent configuration on your leases, listings, policies, and scripts
- Integration work against the systems named in Design
- Guardrails: refusal behaviour, escalation triggers, approval steps
- Testing with a sample of your historical cases and edge cases
- Team walkthrough so staff know what the AI will and will not do
What you receive
- Working automation in staging, then in production behind your rules
- Runbook covering monitoring, overrides, and manual fallback
- Handover session with your team and written configuration notes
AutomateKeep it working as your policies, staff, and portfolio change
Once live, the system needs attention: conversations to review, edge cases to fold in, integrations that change under you. Automate is the ongoing operations phase — monitoring, monthly reporting against the baseline, and iteration on what the logs show.
What we need from you
A short monthly review call and a point of contact for policy changes that affect the AI.
What happens
- Monitoring and alerting on failures, timeouts, and escalation spikes
- Review of real conversations and automation runs for gaps
- Monthly iteration: new cases, tone corrections, workflow changes
- Reporting against the metrics agreed in Design
- Planning the next workflow once the current one is stable
What you receive
- Monthly report on the agreed metrics versus baseline
- Change log of configuration and workflow updates
- Recommendation on the next automation worth building
Four rules the process is built around
These are the decisions that keep an automation programme from becoming an expensive experiment.
Small first release
The first thing we ship is deliberately narrow: one workflow, one integration, one measurable metric. It is easier to trust an automation you have watched work for a month.
Human checkpoints by default
Anything with financial, legal, or tenant-relationship consequences is prepared by the automation and approved by a person. Silent irreversible actions are a design failure.
Your baseline, not our benchmarks
We record your current response times and hours before launch and report against those. Industry averages are not evidence that your operation improved.
Documented, not tribal
Every automation ships with a runbook: what triggers it, what it touches, how to pause it, and who to call. You are not dependent on one person's memory.
If a workflow is not written down, it is not ready to automate. The first thing we build with you is a description of how the work actually happens today.
Two ways to carry on
Read how work is scoped and priced, or look at the scenarios we build most often.
Frequently asked questions
What happens in the Discover step?
What do I receive from the Design step?
Do you stop after Build?
How involved does my team need to be?
See what AI can actually automate in your business
Book a free 30-minute AI Opportunity Audit. We map your current workflows, name the two or three that AI can carry, and tell you plainly where it would not help.