Multi-Company Panel Discussion on Enterprise AI and Agentic AI Deployment Challenges

glean_/deloitte/_docusign · Tech · 2025 · source Agentic: yes · promotional bias: med · source-backed · extracted with prompt 0.5.6 (schema v2)

Generated from extraction.json — do not edit by hand. Every claim cites verbatim source evidence; stated means the quote directly asserts the claim, implied means it supports it indirectly.

Problem

Domain: Enterprise AI deployment

The panel examines why enterprise agentic AI adoption is constrained less by model choice than by organizational complexity, data silos, governance, value measurement, and trust.

Constraints:

Enterprise data and tools are split across silos.
Autonomous agents require governance, human oversight, and security integration.
AI productivity gains are difficult to measure consistently.
Business stakeholders, data teams, technology teams, legal, and security all affect deployment.

Architecture

The panel converges on governed enterprise agent deployment: cross-functional workflow redesign, day-one security involvement, data-silo bridging through connectors or MCP-style APIs, evaluation in production, and human approval before higher-autonomy operations.

Diagram confidence: reconstructed (reconstructed from prose).

flowchart LR
  business-workflow-owners["Business Workflow Owners"]
  ai-agents["AI Agents"]
  enterprise-data-connectors["Enterprise Data Connectors"]
  security-and-governance-teams["Security and Governance Teams"]
  human-reviewers["Human Reviewers"]
  evaluation-layer["Evaluation Layer"]
  business-workflow-owners --> ai-agents
  business-workflow-owners --> security-and-governance-teams
  ai-agents --> enterprise-data-connectors
  ai-agents --> human-reviewers
  ai-agents --> evaluation-layer
  enterprise-data-connectors --> ai-agents
  enterprise-data-connectors --> security-and-governance-teams
  security-and-governance-teams --> enterprise-data-connectors
  security-and-governance-teams --> ai-agents
  human-reviewers --> ai-agents
  human-reviewers --> evaluation-layer
  evaluation-layer --> ai-agents
  evaluation-layer --> human-reviewers

Drag to pan · scroll or use +/− to zoom · reconstructed from the source's described component topology

Components

Component	Role	Interacts with
`Business Workflow Owners`	Define value, success criteria, and redesigned workflows.	`AI Agents`, `Security and Governance Teams`
`AI Agents`	Assist or automate enterprise workflows under evaluation and oversight.	`Enterprise Data Connectors`, `Human Reviewers`, `Evaluation Layer`
`Enterprise Data Connectors`	Bridge silos through connectors, MCP, and wrapper APIs around legacy systems.	`AI Agents`, `Security and Governance Teams`
`Security and Governance Teams`	Approve access, policies, and sharing controls from the start.	`Enterprise Data Connectors`, `AI Agents`
`Human Reviewers`	Approve and monitor agent outputs before full autonomy.	`AI Agents`, `Evaluation Layer`
`Evaluation Layer`	Measures quality, productivity, and readiness for autonomy.	`AI Agents`, `Human Reviewers`

Data flow

Enterprise workflows identify business value first, connect relevant data sources through connectors or APIs, expose context to AI systems under security controls, evaluate outputs, and keep humans in approval loops until trust improves.

“The panelists explore challenges around organizational complexity, data silos, governance, agent creation and sharing, value measurement, and the tension between autonomous capabilities and human oversight.”summary, stated

“better enterprise context leads to better AI responses, and that bridging data silos is key to success.”summary, stated

Why this architecture

The panel argues that enterprise AI succeeds when workflows are reimagined with stakeholders, data access, security, and value capture together rather than by bolting an LLM onto a legacy process.

“successful AI implementation, particularly agentic AI, isn't about "bolting an AI solution to a legacy scenario" but rather "reimagining workflows and processes."”summary, stated

“business value should drive technical decisions rather than focusing solely on which LLM model to use.”summary, stated

Models used: GPT models · Gemini models · enterprise search models

Scale

The panel cites a Gartner prediction that 33% of enterprise apps will include agentic AI and enable 15% of enterprise decisions.
“referencing a Gartner prediction that 33% of enterprise software applications will include agentic AI enabling 15% of enterprise decisions.”summary, stated
Deloitte reports AI activity across life sciences, financial services, banking, and insurance.
“Rohit from Deloitte confirms AI is pervasive across industries including life sciences, financial services, banking, and insurance”summary, stated

Alternatives considered (roads not taken)

Bolting AI onto legacy scenarios

Rejected because: The panel says agentic AI requires workflow and process redesign.

“isn't about "bolting an AI solution to a legacy scenario" but rather "reimagining workflows and processes."”summary, stated

Fully autonomous agents immediately

Rejected because: Most enterprise use cases still require monitoring and validation before actions.

“the current state across most enterprises involves careful monitoring and validation of agent outputs before allowing them to execute actions.”summary, stated

Patterns

`human-in-the-loop-gate` (governance)

The panel treats human approval and validation as the current default for enterprise agents.

“he emphasizes that "human in the loop is critical" and that evaluation mechanisms must be in place not just during pilot phases but also in production deployments.”summary, stated

`safety-policy-enforcement` (governance)

Security teams are involved from day one to handle access and context sharing.

“security teams should be engaged "from day one" to navigate access challenges across these data silos and enable context sharing with AI systems.”summary, stated

`context-engineering` (context)

The architecture centers on bridging enterprise context from silos into governed AI systems.

“He stresses that better enterprise context leads to better AI responses, and that bridging data silos is key to success.”summary, stated

Tradeoffs

Keep humans in the loop

Chose Monitored and approved agent outputs over Immediate full autonomy.

Rationale: Enterprise trust is built incrementally through evaluation and consistent performance.

“as agent pipelines mature and organizations gain confidence through consistent performance, they may gradually move toward more autonomous operation.”summary, stated

Start from business value

Chose Roadmaps and success criteria defined upfront over Model-first experimentation.

Rationale: Business value should determine technical choices.

“Deloitte now emphasizes strategic approaches to roadmaps, business value capture defined upfront, clear success criteria, and comprehensive stakeholder engagement from the beginning.”summary, stated

Gotchas (what bit them)

Organizational complexity mirrors technical complexity.

Root cause: Enterprise tools, data silos, approval layers, and training processes evolve more slowly than models.

Resolution: Use cross-functional roadmaps, security involvement, and success criteria from the beginning.

“While models evolve quickly, organizational processes around training, security approval, and legal review move more slowly, creating friction.”summary, stated

Independent experimentation creates tool proliferation.

Root cause: Teams choose different tools for similar problems without shared governance.

Resolution: Move from individual experimentation to governed enterprise-wide agent deployment.

“different teams selecting different tools for similar problems, confusion about when to use which models”summary, stated

Evaluation

The panel emphasizes production evaluation and business-value measurement but does not report a single benchmark.

Metrics: productivity gains · business value · agent performance consistency

“the difficulty of measuring AI-driven productivity gains”summary, stated

Outcomes

The panel identifies governance, data integration, workflow redesign, and human oversight as enterprise AI deployment requirements.
It frames autonomous operation as aspirational for most enterprise use cases.
It argues business value should drive technical decisions.

The discussion covers multiple enterprise perspectives, but does not report a single deployment metric.

“The conversation, hosted at AWS re:Invent, offers candid insights into the operational realities of deploying large language models and agentic AI systems in production environments.”summary, stated