AI AI Workforce OS v1.0.0

AI Platform

AI dashboard, providers, prompts, agents, skills, and usage analytics.

System AdministratorsHR Administrators

Overview#

The AI Platform connects AI Workforce OS to large language model providers and exposes that capability through managed prompts, autonomous agents with an approval workflow, a fixed set of HR-focused "AI Skills," and usage/cost tracking.

Purpose#

Explain what the AI Platform actually does today, since "AI" covers several distinct pieces (providers, prompts, agents, skills) that are easy to conflate.

Prerequisites#

  • A System Administrator to configure at least one AI provider with valid API credentials.
  • Outbound network access from your server to the configured provider (OpenAI, Anthropic, or Google Gemini).

Step-by-Step Instructions#

Configuring a provider and model#

  1. Go to AI → Providers and add a provider — supported providers connect to OpenAI, Anthropic, or Google Gemini.
  2. Add the specific model(s) you want available under that provider.
  3. Set which provider/model is used as the default where a skill or agent does not specify one.
ⓘ
Verify connectivity in your environment

Provider integrations call each provider’s real API over HTTPS. Confirm connectivity and valid credentials in your own environment before relying on this in production — test with a simple prompt first.

Using AI Skills#

  1. Go to AI → Skills to see the available skills: resume analysis, job description generation, interview question generation, interview feedback summary, goal generation, KPI generation, review summary, learning path/course recommendation, skill-gap analysis, HR policy Q&A, document summarization, employee profile summary, payroll explanation, attendance insight, leave recommendation, and asset recommendation.
  2. Run a skill against the relevant record (for example, a candidate’s resume, or an employee’s profile).
  3. Review the AI-generated output before acting on it — skills are assistive prompts, not automated decision-makers.
ⓘ
Fixed skills, not a builder

AI Skills are a fixed, pre-built set of prompt-driven endpoints. There is no user-authorable "build your own AI skill" feature in this release.

Working with prompts#

  1. Go to AI → Prompts to view, create, or update prompt templates.
  2. Prompt templates are versioned — each edit can be saved as a new version.
  3. Review conversation history under AI → Conversations (view-only).

Agents and approvals#

  1. Go to AI → Agents to see registered agents and run one against a target.
  2. Review a specific agent run and its individual steps for transparency into what the agent did.
  3. Some agent actions require human approval — review pending approvals under AI → Agents → Approvals and decide (approve/reject) before the action proceeds.

Insights, recommendations, and search#

  1. Review AI-generated executive insights and recommendations where surfaced in the AI dashboard.
  2. Use natural-language search to query the system in plain language instead of building filters manually.

Usage and cost#

  1. Go to AI → Usage to review request volume, token usage, and cost, broken down by provider/model.

Screens Involved#

📷
[Screenshot: AI dashboard]
📷
[Screenshot: Providers / Models]
📷
[Screenshot: Prompts]
📷
[Screenshot: Conversations]
📷
[Screenshot: Agents / Agent runs]
📷
[Screenshot: Agent approvals]
📷
[Screenshot: Skills]
📷
[Screenshot: Search]
📷
[Screenshot: Usage (cost, requests, tokens)]

Expected Results#

Configured providers respond to requests, AI Skills produce useful drafts for a human to review, and agent actions that need approval wait for a human decision before proceeding. Usage and cost are visible per provider.

Not Available in This Release#

🚫
Not currently available
  • No model fine-tuning.
  • No vector database, embeddings, or retrieval-augmented generation (RAG) pipeline.
  • No image or audio generation.
  • No third-party AI marketplace or plugin ecosystem.
  • No user-authorable custom-prompt "app builder" — AI Skills are a fixed set defined by the product.

Common Mistakes#

  • Treating AI Skill output (e.g., a review summary or resume analysis) as final without human review.
  • Assuming "attrition risk" or similar analytics elsewhere in the product are AI-generated — check the specific module’s documentation; not everything labeled "analytics" runs through the AI Platform.
  • Not configuring provider credentials, then reporting AI features as "broken" when they are simply unconfigured.

Troubleshooting#

An AI Skill returns an error.#

Confirm a default provider/model is configured with valid credentials, and that your server has outbound network access to that provider. Check AI → Usage for failed-request indicators.

An agent action seems stuck.#

Check AI → Agents → Approvals — the action may be waiting on a human decision before it can proceed.

RecruitmentPerformanceLearning

FAQ#

Which AI providers are supported?

OpenAI, Anthropic, and Google Gemini, each configured with your own API credentials.

Can I create my own AI skill or prompt-driven feature?

You can create and version prompt templates under AI → Prompts. The fixed set of "AI Skills" (resume analysis, review summaries, etc.) is defined by the product and is not user-extensible in this release.

Do AI agents take actions automatically without oversight?

Some agent actions require explicit human approval before they proceed — check AI → Agents → Approvals. Not every action is gated this way, so review your agent configuration.

Best Practices#

  • Always have a human review AI-generated content (summaries, recommendations, generated text) before it is used in an official record or sent externally.
  • Monitor AI → Usage regularly if you are on metered provider billing, to avoid surprise costs.