Home / Guides and tools / Ground truth wins
Give AI a record it can work from
An assistant can draft a convincing answer from incomplete records. In a services firm, that matters when the sold scope, the live work and the billing decision sit in different places. Before you add an AI task to the process, ask whether your own team can trace those decisions.
Follow one client commitment
Take a recent engagement and find four things. If an owner has to rebuild any of them from email or memory, an assistant will face the same gap.
| Link | What to find | If it is missing |
|---|---|---|
| Approved scope and rates | The version the client agreed, not the one in the proposal draft. | Delivery plans from a version nobody can confirm. |
| The decision that changed them | A change, who approved it and when. | The change lives in an email thread. |
| The work the team recorded | Time and costs tied to that scope. | Hours sit against a task nobody can map back. |
| The invoice finance prepared | Each line traced to the work behind it. | Finance rebuilds the invoice by hand. |
A useful AI workflow needs two things: a record to read, and a controlled path for proposing a change.
Three questions before any AI task
Ask them of any assistant, from any vendor, against the engagement you just traced.
What can it see?
The records it reads. If the scope, rates, changes and invoice are not connected, it sees fragments.
What can it propose?
A draft change for a person to review. A proposal is not a saved change.
Who checks the result?
A named person, before anything is saved, with a way to correct a wrong result afterwards.
Could a person trace it? Then a tool could.
Four questions about one recent engagement. Nothing is stored or sent.
Where DAY ONE fits
DAY ONE keeps the scope, rates, changes, time and invoice lines on one connected record, so the trace above is a short walk and not a search. On Salesforce; works with your Salesforce org, or available as a bundled edition if you don't use Salesforce.
For Salesforce account teams: a services customer may run its pipeline in Salesforce while delivery and billing happen elsewhere. Ask what becomes of an approved opportunity, and where finance checks the work before it is billed. That conversation finds a real operating need before anyone picks an AI use case.
Questions about the guide
Can't see your question? Ask us directly.
Is the guide an AI product comparison?
No. It is a way to inspect the records, permissions and decisions behind an AI task, before you choose one.
Does a connected record mean every action can be automated?
No. The team still needs to define which actions are allowed, who reviews them and how a wrong result is corrected.
Do I have to give my details to read it?
No. The guide is on this page in full, with nothing to fill in.
Who is it for?
Salesforce account teams and the leaders of services firms who are being asked to add AI to delivery and billing. It is written to be read in one sitting.
Talk through your own engagement
Tell us which engagement decision is hardest to find, and we will talk it through.