> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fincelo.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent Enhancement Roadmap

> 34 planned capabilities across all 7 agents.

### ROADMAP — Q\&A

***

**Q: What are the 5 levels of agent enhancements?**

Level 1 — Smarter agents (8 capabilities, build now):
Deeper reasoning on Fincelo's own data. No new integrations required.
Revenue quality score, payment date prediction, agent outcome learning,
ICP engine, revenue leakage detection, CLV scoring,
Z-score rolling baseline, renewal health composite score.

Level 2 — External signal integration (5 capabilities, Phase 4):
Agents that see beyond Fincelo's database.
India gov portal monitoring (GSTIN + MCA + MSME Samadhaan),
decision-maker departure alert, regulatory change detection,
FX rate hedging signal, competitor signal detection.

Level 3 — Autonomous action within guardrails (8 capabilities, Phase 4–5):
CFO defines parameters upfront. Agents execute within them.
Autonomous deal parameters, document drafting, audit evidence packer,
voice command → action, agent conflict prevention,
customer success playbooks, deal desk automation, NPS pipeline.

Level 4 — Predictive intelligence (7 capabilities, Phase 5, needs 6+ months data):
ARR target probability (P10/P50/P90), predictive period close,
cash runway, pricing intelligence, revenue forecast bands,
multi-entity treasury view, predictive churn.

Level 5 — Fully autonomous finance intelligence (6 capabilities, 2027):
Multi-agent debate, industry benchmarking network,
autonomous period close, self-healing revenue ops,
board report auto-generation, investor data room update.

***

**Q: What is agent outcome learning?**

Level 1 enhancement across all agents. Each agent records the outcome of every
action it takes (did the dunning email lead to payment? did the renewal alert
lead to quote sent?).

Over time, the agent learns which actions work for which customer types.
Low-efficacy actions are deprioritised. High-efficacy actions are weighted higher.
This is the compounding moat — after 6 months on Fincelo, the agents know what works
for your specific customer base better than any generic playbook.

***

**Q: What is the token governance system for agents?**

Agent actions consume credits from the workspace's monthly allocation:

| Plan            | Monthly allocation  | Per-agent weight       |
| --------------- | ------------------- | ---------------------- |
| Growth + Agents | 1,200 actions/month | 1–3 credits per action |
| Enterprise      | 5,000 actions/month | 1–3 credits per action |

Credit weights by action type:

* 1 credit: read/analyse (health score, briefing generation)
* 2 credits: communicate (send email, create task)
* 3 credits: write/execute (post journal entry, update CRM, suspend account)

Alerts at 50% / 70% / 90% / 100% of monthly allocation.
10% grace buffer above limit before hard stop.
Overage opt-in available per tier.

***