Hermes enriches CRM exports with confidence columns
Noronha reports a daily Hermes workflow that fills missing CRM-export fields into a sourced confidence-labelled CSV, leaving uncertain matches for human review and import.
Original prompt
Shared by the source author. See source notes for context.
Run the enrich-crm-record skill. Read the latest contacts export CSV from the watched folder.
Reach people exploring AI agents
See placements →What happened
Noronha reports a daily Hermes workflow that fills missing CRM-export fields into a sourced confidence-labelled CSV, leaving uncertain matches for human review and import.
Reported by Reuben NoronhaHow the agent did it
- Export incomplete CRM contacts into the watched folder.
- Define required fields and confidence rules before research.
- Research identity and missing fields, recording a source per value.
Where a person stepped in
A human checks and imports the output CSV. Low-confidence matches remain blank; the source does not establish a measured production accuracy rate.
Reuben Noronha uses Hermes to enrich incomplete CRM records without granting it direct write access to the live CRM. A scheduled export places contacts in a watched folder. Hermes identifies missing required fields, researches the person and company, reconciles conflicting information and writes a separate output CSV.
The required fields include job title, company, employee-size band, industry and LinkedIn URL. Each filled value has an adjacent source. Direct signup-profile URLs or exact email-domain matches support high confidence; a name-and-company secondary match receives medium confidence. A name-only match with multiple plausible candidates stays blank and is marked for review.
The scheduled instruction caps a run at fifty rows and emails a summary with the file. A person scans the result and imports it into the CRM. The article's sample counts illustrate the report format, rather than establishing an audited production total or accuracy rate. The useful boundary is the separate CSV: identity uncertainty remains visible before anyone updates live sales data. Noronha recommends spot-checking enriched rows during the initial weeks and retaining provenance, since otherwise an incorrect title or employer can silently propagate into segmentation and outreach.
Prompt
Original prompt
Source: Blog · Reuben Noronha
Frequently asked questions
Why keep CRM writes manual?
The separate output CSV provides a human checkpoint before uncertain research affects live contact records.
What happens to ambiguous identities?
Name-only matches with multiple plausible people remain blank and receive a needs-review mark.
What makes a filled field auditable?
An adjacent source column records where the value came from, alongside the confidence tier.
More details
Remaining steps
- Leave low-confidence matches blank and prepare the output CSV.
- Have a person inspect and import the reviewed file.
About this account
Scheduled Hermes runs inspect contact rows, research and reconcile missing data, then produce a confidence-labelled file and email summary without live CRM writes.
Firsthand account by Reuben Noronha; original publication date verified against Primary blog byline July 3, 2026. Outcomes are the author’s reports, not independently reproduced. Original scheduled-task excerpt; the full instruction includes a 50-row cap and confidence rules.
Credit: Reuben Noronha
Source published 3 Jul 2026
Source checked 14 Oct 2026
Tools and topics
Hermes cron · CRM export CSV · Watched folder · Web search · Enrichment sources · Email
CRM completeness · Identity matching · Source columns
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.