Traditional CRM reporting often fails to answer the most critical question in sales management: why are we missing revenue? TeamPulse AI configuration in SuiteCRM solves this by providing a dual-layer behavioral intelligence architecture – a deterministic built-in engine and an optional generative AI layer – that translates complex activity patterns into plain-English coaching narratives.
Whether using the always-on InsightEngine or a large language model (LLM) via BYOK, TeamPulse ensures coaching is proactive, data-backed, and focused on leading behavioral indicators rather than lagging financial outcomes.
The Two-Layer AI Architecture
| Feature | InsightEngine (Built-in) | Generative AI (BYOK) |
|---|---|---|
| Requirement | Included out-of-the-box. No API key required. | Requires valid API key (OpenAI, Gemini, Azure, Ollama). |
| Technology | Deterministic, rule-based analysis. | Large Language Model (LLM) processing. |
| Internet | Operates locally on your SuiteCRM server. | Requires outbound connection to AI provider. |
| Output Type | Activity anomalies, severity-ranked alerts. | Natural-language narratives, coaching scripts. |
| Primary Use | Core behavioral signals and verdicts. | Deep reasoning and comparative analysis. |
Built-in InsightEngine
The InsightEngine is an always-on, rule-based system that requires no external configuration. It provides:
- Activity Anomaly Detection – Identifies sudden spikes or drops against a rep’s historical baseline.
- Pattern Detection – Recognizes specific behavioral archetypes such as the “Call Machine” (high volume, low conversion) or the “Deal Staller” (high open deals, zero activity).
- Coaching Signal Generation – Produces structured alerts based on deal and activity patterns.
- Deterministic Verdicts – Assigns team health levels (Excellent, Growing, Stable, Declining, or Critical).
Generative AI (BYOK – Bring Your Own Key)
For organizations seeking richer, more nuanced analysis, TeamPulse supports external LLM integration. This layer adds human-like reasoning to the data:
- Natural Language Summaries – A synthesized “Why” behind a rep’s struggle or success.
- Coaching Scripts – Suggested talking points for 1:1 meetings based on specific deal gaps.
- Comparative Analysis – Deep-dive comparisons between a rep’s performance and the team’s top performers.
Supported AI Providers
- OpenAI – GPT-4o, GPT-4o Mini (recommended), and GPT-4 Turbo.
- Google Gemini – Gemini 2.0 Flash (recommended), 1.5 Flash, and 1.5 Pro.
- Azure OpenAI – For enterprise-grade, private cloud deployments.
- Ollama (Local) – Supports Llama 3.2, Mistral, and other models for organizations requiring on-premises AI processing with no external data transmission.
Step-by-Step AI Setup
AI settings take effect on the next summary generation cycle, which runs every 30 minutes via the SuiteCRM scheduler (“TeamPulse – AI Summary & Cache Refresh”).
- Navigate to Settings – Access TeamPulse from the SuiteCRM sidebar and click the Settings tab.
- Open AI & Data – Select the AI & Data sub-tab.
- Select Provider – Choose your preferred AI provider from the dropdown. If you do not have an API key, select InsightEngine (built-in).
- Enter API Credentials:
- For OpenAI or Gemini – Paste your API key. Keys are stored with AES-256 encryption in the
teampulse_configtable. - For Azure OpenAI – Provide the Endpoint URL and Deployment Name.
- For Ollama – Enter your local server URL (default:
http://localhost:11434).
- For OpenAI or Gemini – Paste your API key. Keys are stored with AES-256 encryption in the
- Choose a Model – Select a specific model (e.g.,
gpt-4o-miniorgemini-2.0-flash). - Customize Prompts (Optional) – Add custom instructions to guide the AI’s tone or focus (e.g., “Refer to reps by first name only”).
- Set Cache TTL – Define how long summaries are stored (default: 6 hours) to minimize API costs.
- Save and Test – Click Save AI Settings, then use the Test Connection button to verify the configuration is reaching the provider.
[Image Placeholder: Settings AI and Data tab showing provider dropdown, API key field, model selector, and Test Connection button]
AI Safety Guardrails and Privacy
TeamPulse is engineered with strict guardrails to keep focus on behavioral coaching:
- No Financial Forecasting – The AI is strictly prohibited from guessing revenue numbers or financial impact. It manages behavior, not dollar outcomes.
- Zero Data Retention – TeamPulse does not store your data for training purposes. Requests are sent directly from your server to the provider.
- Behavioral Only – All prompts are limited to behavioral signals (calls, tasks, deal hygiene) to prevent hallucinations regarding employment status or financial advice.
Next Steps After AI Configuration
- Send Targeted Coaching Notes – Use the Action Required panel in the AI Summary to identify which reps need immediate 1:1 intervention.
- Draft Journal Entries – Use the AI Draft button in the Coaching Journal to pre-fill performance milestones based on AI-detected patterns.
- Intervene on Stale Deals – Review Deal Hygiene flags highlighted by the AI to identify pipeline rot before it affects the forecast.
- Public Recognition – Identify “Momentum Leaders” for praise during team meetings.
Troubleshooting / FAQ
Q: Why does the AI Summary say “No significant data”?
A: The AI requires at least 5-7 days of behavioral history to generate a meaningful narrative. Ensure reps have logged activities (calls, meetings, tasks) within the selected date range.
Q: Is my API key safe?
A: Yes. API keys are stored as AES-256 ciphertext in the teampulse_config table. They are only decrypted server-side when making a direct call to the configured AI provider.
Q: Can I use AI without an internet connection?
A: Yes. Select InsightEngine (built-in) and TeamPulse performs all analysis locally on your SuiteCRM server using deterministic logic, with no external API calls required. Alternatively, configure Ollama for a local LLM.