by Bryon Bratcher, Managing Director, Gravity Stack
At Gravity Stack, we’ve been leaning into AI since our launch in 2018, not just in theory, but practically. Over the last couple of years (2023-2024), generative AI assistants have significantly improved how we perform research, brainstorm creatively, and develop strategies. These tools have been nothing short of life-changing for the way our team is working.
But now, we’ve shifted our attention to AI agents. The term “AI agent” often gets tossed around loosely, so let’s clarify exactly what we mean, especially when compared with assistants:
- AI Assistants: Tools that support humans in discrete tasks, such as research, summarizing, or creative idea generation.
- AI Agents: Autonomous systems designed to manage entire complex workflows dynamically—handling adjustments, revisiting steps when necessary, and adapting in real-time without constant human oversight.
Here’s why this distinction matters in practice. First, because as much as assistants can improve productivity, they’re not designed to automate workflows. They are designed to help a single individual work through challenging tasks.
Of course, automation did not begin with AI agents, but traditional legacy automation typically operates in a linear way. Consider a salesperson submitting a non-standard contract through a legacy Contract Lifecycle Management (CLM) system. If the contract contains unexpected terms or ambiguities, the workflow stalls, demanding manual intervention and potentially restarting the entire process.
In contrast, an AI agent handles the same CLM scenario dynamically. If unexpected terms or ambiguities arise, the agent doesn’t stall—it loops back and adapts.
You can see this difference in the following execution log for an agentic workflow running an M&A transaction.
[INFO] Received instruction document: "M&A Due Diligence for Acquisition of XYZ Corp…."
[INFO] Extracting key focus areas from the instruction document...
[PLANNING AGENT] Key priorities identified:
- Verify intellectual property (IP) ownership.
- Check indemnification and liability clauses.
- Identify change-of-control risks.
- Highlight exclusivity restrictions.
[PLANNING AGENT] Plan generated successfully.
[EXECUTION AGENT] Processing Document 1...
[EXECUTION AGENT] Extracting key entities...
→ Identified: Party A, Party B, Contract Type: Licensing Agreement
[EXECUTION AGENT] Checking IP clauses...
→ Clause Found: ✅ "Party A retains all IP rights post-termination."
[EXECUTION AGENT] Checking indemnification...
→ Unexpected issue: Clause references an outdated law! ⚠️
→ Self-reflection: "Should I flag this or check another source?" → Decision: Flagging for review.
[EXECUTION AGENT] Change-of-control risks...
→ Clause Found: ⚠️ "Contract automatically terminates upon ownership change."
[EXECUTION AGENT] Summary: Moderate Risk ⚠️ (Outdated indemnification clause, potential contract termination).
[EXECUTION AGENT] Processing Document 2...
[EXECUTION AGENT] Extracting key entities...
→ Identified: Party C, Party D, Contract Type: Supplier Agreement
[EXECUTION AGENT] Checking indemnification...
→ ❌ No indemnification clause found.
→ Self-reflection: "Should I check alternative sections?" → Decision: Searching contract appendix... ✅ Found in Appendix A.
[EXECUTION AGENT] Checking exclusivity restrictions...
→ Clause Found: ✅ "Supplier must prioritize Party C for 2 years."
[EXECUTION AGENT] Summary: Low Risk ✅ (All clauses in place after extended search).
[REPORT AGGREGATION AGENT] Analyzing contradictions and risks...
→ 🚨 Inconsistent indemnification clauses across documents.
→ 🚨 Change-of-control clause in Document 1 may void contract.
→ Recommendation: Standardize indemnification language and renegotiate Document 1.
[REPORT AGGREGATION AGENT] Final Risk Assessment: Moderate to High 🚨
[INFO] Due diligence completed. Final report ready for review.
As a matter of fact, we recently met with an organization building agents to manage an entire M&A document review process—going beyond pulling clauses, agents ran the entire workflow dynamically. As a company that has been using AI based diligence tools since 2018, it was nothing short of jaw dropping.
At the same time, GenAI assistants remain valuable, especially for creative strategy, brainstorming, and idea generation. But agents represent a fundamental shift in what automation can achieve. We recently met with a company automating IP litigation workflows using agents. As the entrepreneur put it to us, “We sell work, not efficiency.”
At Gravity Stack, we continuously research and test cutting-edge tools so you don’t have to. If you’re considering AI agents or just want to learn more, reach out.
