Least access
Each agent gets only the permissions its task needs.
Home › How we work
How we work
Small first, then real
Four phases, each with a clear result and a decision point. You see a working agent before you commit to a rollout, and nothing important happens without your approval.
The four phases
Discover → Pilot → Build → Run
Tap a phase to see what happens and what you get.
01 · Discover
We learn how the work is done today, where time and money leak, and whether your data is ready.
What happens
- An intro call about your goals
- A review of your workflows, systems and data
- A Salesforce health check, if you use Salesforce
What you get
- A short report of what to fix and automate first
- One recommended first use case
- A one-page scope with a fixed price
02 · Pilot
A working agent on one real workflow, so you see results before committing to more.
What happens
- Build on sample data, then a safe slice of real data
- Weekly check-ins with your team
- Measure against the result we agreed
What you get
- A working agent with approval steps and logging
- Before-and-after measurement
- A clear recommendation: roll out, adjust or stop
03 · Build
We take what worked in the pilot into production, alongside your existing systems and automation.
What happens
- Design reviewed with your team
- Build in a sandbox, never straight in production
- Testing with real scenarios before go-live
What you get
- Agents and integrations in production
- Permissions, approvals and audit logs
- Documentation your team can run
04 · Run
We keep everything working and improving, or hand it over to your team. Your choice.
What happens
- Monitoring, fixes and tuning
- Monthly improvements and new agents
- Salesforce admin and development, if you need it
What you get
- A named lead who knows your setup
- Every change tested and documented
- Cancel any time with notice. No lock-in
Guardrails
Autonomous, not unsupervised
Human approval
Sending, paying, deleting and record changes wait for a person.
Every step logged
You can see what each agent did, when and why.
Tested first
Sandbox and sample data before anything touches production.
Data & security
Your data stays yours
NDA first
We sign before we see anything.
Named access
Our own named logins, never shared accounts, removed when we finish.
Minimal data
Agents read only what a task needs, through access you control.
Agreed in writing
Data handling is part of the agreement before any build starts.
What we need
What we need from you
Small asks, big difference.
- One person who knows the workflow well
- A test or sandbox environment where possible
- Sample data to build and test with
- About an hour a week for review and approvals
How long does each phase take?
It depends on the workflow and your systems. We agree the timeline in the scope before any work starts, and stick to it.
Can we stop after the pilot?
Yes. The pilot exists so you can decide on real results. You keep what was built.
Where does our data go?
As little as possible leaves your systems. Agents read only what a task needs, through access you control, and we agree data handling in writing before any build starts.
Do you sign NDAs?
Yes, before we see anything.
Who owns the work?
You do. Code, configuration and documentation are handed over. No lock-in.
Start a conversation
What repetitive work would you hand to an agent?
hello@nodepointai.com