Skip to main content
Once PG:AI is connected, use natural language. These examples show what the model typically does behind the scenes - you do not need to name tools yourself.

Account research

Profile and insights
Get the SWOT analysis and strategic priorities for Salesforce.
Uses company with include: ["swot_analysis", "strategic_priorities"]. Deep dive
Give me the full company profile for Tesco - overview, competitors, events, and value pyramid.
Uses company with include: ["full"] or multiple sections. Semantic search
Search my workspace for anything about cloud migration at Frasers Group.
Uses search with scope: "organisation" and a company filter if needed. Jobs
What data engineering roles is Harness hiring for in the UK?
Uses jobs_search with q, company, and country_code.

Accounts and discovery

List your book
List my tracked accounts.
Uses accounts with default scope: "user". Org-wide view
Show all accounts in the organisation that mention “fintech”.
Uses accounts with scope: "org" and query. Find new companies
Search globally for companies named Snowflake - tell me if we already track them.
Uses companies_search - check is_in_user_accounts in the response. Add to workspace
Add stripe.com to my tracked accounts.
Uses companies_add, then get_operation_status until enrichment completes.

Contacts

Find by role
Find VP of Engineering contacts at Salesforce.
Uses contact_search with title criteria and company scope. Full profile
Get full details for contact ID abc-123-….
Uses contact (after search returns an ID). Save a contact
Save Jane Doe, Head of Data at Tesco - [email protected] - to my workspace.
Uses create_contact. Enrich email
Find the work email and mobile for contact abc-123-….
Uses contact_find_email_phone, then polls get_operation_status, then contact.

Content generation (full flow)

Single document (e.g. value pyramid)

You:
I want to generate a value pyramid for Salesforce. What templates are available?
AI: Calls templates with produces: "document" (or unfiltered), shows options. You:
Use the Value Pyramid template for Salesforce. Focus on data platform consolidation.
AI:
  1. template_generate - template_kind: "canvas", template_id from step 1, company: "Salesforce", user_instruction with your focus
  2. get_operation_status - poll until completed
  3. canvas_get - format: "text" for readable output

Email or call script

List email templates, then generate a professional email to the CIO at Airtable about analytics governance.
Same flow - pick a canvas template where produces is email or call_script.

Outreach sequence

Show sequence templates, then generate a 5-step email sequence for Harness aimed at platform engineering leaders.
  1. templates with produces: "sequence"
  2. template_generate with template_kind: "sequence"
  3. get_operation_status
  4. sequence_get

Browse existing content

List documents
What value pyramids have we already generated for Salesforce?
Uses canvas_list with company and content_type: "value_pyramid". Read one
Open canvas 98b4c290-… as plain text.
Uses canvas_get with format: "text". Sequences
List draft sequences for my territory.
Uses sequence_list with status: "draft".

Agent conversations (advanced)

Start a company-scoped agent chat about Microsoft - ask what changed in their earnings narrative this quarter.
Uses agent_message_send with company: "Microsoft", then get_operation_status.
List my recent agent sessions for Airtable.
Uses agent_sessions_list.

Multi-step conversation

A natural research → generate chain:

Tips for better results

Say which angles you care about: “SWOT and competitors” rather than “everything about them” unless you want include: ["full"].
Names are fuzzy-matched against tracked accounts. Domains (stripe.com) also work.
After generation or company add, ask the model to “check operation status” if it stops early.
Ask for “templates I can use to generate” vs “canvases we already created” - different tools.
Adding companies and contact enrichment use credits. If you get a 402 error, check your organisation credit balance in PG:AI.