# Sumit Johar

Sumit Johar leads IT and security at BlackLine and writes about enterprise AI, with a focus on governance, automation, and financial operations.

- Profile: https://b2binfluencer.marketing/creators/sumit-johar/
- LinkedIn: https://www.linkedin.com/in/sumit-johar
- LinkedIn followers: 6,985
- Topics: artificial-intelligence, leadership, ai-governance
- Based on 15 dated posts, latest 2026-09-17

## Summary

Sumit Johar is useful for thought leadership on AI governance, agent controls, and finance automation.

- Best for: A B2B SaaS AI governance and finance automation campaign for CFO, CIO, and finance operations audiences.
- Not for: A developer tools campaign for hands-on machine learning platform engineers wanting tutorials, code walkthroughs, and demos.
- Check first: Confirm audience job function, seniority, region, engagement, media kit, rates, deliverables, approval, disclosure, and usage terms.

### Takeaways

- Every AI agent needs a governed control plane between agent and systems of record.
- Finance workflows need policy checks, dual approval, segregation of duties, and audit logs.
- AI ROI must be entry ticket, not post hoc justification.

## Main ideas

- **Build or buy is the wrong question — the real one is what every agent runs through**: Johar argues that treating AI agents as one-off build-or-buy decisions misses the real risk: what every agent passes through before it touches a system of record. They describe three layers: the agent layer, where the intelligence lives; the systems of record, where transactions live; and a control plane in between, which enforces identity, data lineage, policy, and an audit trail recording what an agent did, when, and on whose authority. They write from firsthand experience running IT and security for a company whose entire product is financial trust, seeing the problem from both sides of the table. The alternative to one governed control plane, they warn, is fifty ungoverned integrations a team quietly owns forever — shadow AI with a change ticket attached.
- **AI governance means different things to a CISO, a data leader, and a CFO**: Johar argues that "AI governance" means different things to different people. A CISO sees cybersecurity: data privacy, and preventing prompt injection and data leakage. A data leader wants reliable data outcomes with the right access controls. A CFO or controller wants to guard the general ledger, enforce internal controls, and guarantee auditability. The proposal is to pair probabilistic models with domain-specific policy enforcement, so automated work still stops at deterministic policy before affecting financial records.
- **Build where the advantage is yours to create; buy where it already exists**: Johar frames build-versus-buy as a question of durability, not difficulty. A capable team can stand up an agent in weeks, but production requires governance across hundreds of agents, evaluation suites refreshed as models churn every six to twelve months, and ongoing compliance work. Their filter: build where the workflow offers genuine differentiation — proprietary data, internal processes, or a customer experience nobody can sell you — and buy where the stakes are highest and tolerance for error is zero. Costs that emerge at scale include a governance layer nobody designs on day one, re-validation each time a model is deprecated, and key-person risk when the few engineers who understand the prompt scaffolding leave.
- **AI ROI has to be the entry ticket, not the post-hoc justification**: Responding to a CFO.com piece about enterprise AI pricing models, Johar argues that escalating and unpredictable costs — premium model tiers, token consumption, compute overhead — can quietly erode projected ROI before a rollout finishes. ROI, they argue, should be a condition for funding, not a story told after the fact. They are not arguing for slowing AI down, but for making investments that survive contact with a CFO's spreadsheet.
- **Agents broke the SaaS trust model, and SaaS discipline has to be earned back**: Starting from OpenAI's disclosure of models acting outside their instructions, Johar argues that giving an agent a goal and a set of tools means it will find a route nobody planned for. They then look back at two decades of SaaS dominance: the usual explanation, cost and convenience, misses the bigger reason. SaaS bundled permissions, audit trails, the approved way to run a process, data everyone trusted, and a vendor whose name was on the contract if something broke. Agents built on general AI platforms carry none of that, leaving responsibility unanswered. Johar concludes that no new trust model is needed; one already works. What is needed is AI capability delivered with SaaS discipline and accountability. They note that they work for a SaaS company, and that bundling accountability and actually delivering it with agents are different things that have to be earned.
- **Trust is built in how a CIO handles bad news**: Johar writes that delivering bad news is unavoidable for a CIO and, handled well, can build trust with management and the board. Trust is established not when everything goes perfectly, but in how incidents and hard conversations are handled. Their advice: build a foundation of transparency early so bad news is never a surprise, and stay neutral and constructive rather than defensive. Proactive communication of risks and trade-offs keeps trust intact.

## Work to read

- [Build vs. buy is the wrong question — what does every agent run through?](https://www.linkedin.com/posts/sumit-johar_agenticfinancialoperations-cio-activity-7487518598874177536-abJD)
- [AI governance means different things to a CISO, a data leader, and a CFO](https://www.linkedin.com/posts/sumit-johar_aigovernance-financeai-cfo-activity-7495285844828741632-0lIb)
- [Where building creates durable advantage, and where it just creates durable maintenance](https://www.linkedin.com/posts/sumit-johar_agenticfinance-cio-buildvsbuy-activity-7492988156888301569-JmTi)
- [Why AI agents break the SaaS trust model — and what has to replace it](https://www.linkedin.com/posts/sumit-johar_cio-responsibleai-blackline-activity-7506485820984442880-ksCi)
- [AI ROI has to be the entry ticket, not the post-hoc justification](https://www.linkedin.com/posts/sumit-johar_claude-pricing-raises-new-budgeting-questions-activity-7466169274940686336-MJZ4)
- [Delivering bad news as a CIO, and why hard moments build trust](https://www.linkedin.com/posts/sumit-johar_ciocom-the-voice-of-it-leadership-activity-7469068763821195265-3NA2)

## Brand relationships

- **BlackLine** (employment)
- **CIO.com** (mention)

## Partnerships

The posts focus on thought leadership about enterprise AI governance, controls for agentic systems, and finance automation, often for CIO, CFO, and finance-operations audiences. They support assessing Sumit Johar as a practitioner voice, but describe no partnership, sponsorship, paid-collaboration format, or paid brand collaboration. Before planning a collaboration, ask for their media kit, rates, deliverables, approval process, disclosure practices, and usage terms.

## Alternatives

- Ahikam Kaufman: If the sponsor wants to shape decisions about the running economics of finance agents rather than control-plane design, these two posts offer a concrete fit. They treat tokens as a new unit of enterprise economics, with every prompt, agent action, and context window carrying a cost; frame model choice for each process as a cost decision; and ask what it would take to change direction once financial logic, team process, and audit history are tied to one platform. They also cite a Deloitte finding that 84% of finance leaders expect technology sovereignty to affect investment decisions through FY2027. That makes this a closer match for a vendor selling predictable cost, portability across the CFO’s existing stack, or an exit path from a closed automation platform than for one making a governance-first argument.
- Tammy Coley: If the sponsor wants the adoption side of agentic controls—the case that governed automation depends on clean, business-owned data, and that finance teams must control their data before trusting agents—the customer-visit post names the pattern directly. Talented finance staff are stuck as the human integration layer in spreadsheets or waiting in an IT queue, while finance leaders approach AI deliberately because, in the author’s words, AI without governance, internal controls, and data integrity is unmitigated risk. A second post frames transformation as a continuation of customer-service values learned in a family business. For a campaign aimed at getting the Office of the CFO to implement controls, this is a stronger voice than one focused on agent architecture.
- Sameer Syed, CPA: If the sponsor wants proof from inside a finance organization already running agents in close, revenue, and procurement work, these posts offer a practitioner account rather than a vendor thesis. They list named deployments: month-end close skills that shortened the close by almost two days; a cash and AR command center for leadership reporting; AI across contract review, revenue validation, and commissions; and a target of 40+ use cases by year-end. The product-launch post is the closest fit for a campaign about agentic finance with controls: accrual automation that generates journal entries from purchase data already in the system, ready to post to the ERP, with a human in the loop and a full audit trail on every calculation—a working example of agentic finance with controls attached.

---
AI-assisted summary of public posts. Audience demographics are not available. Engagement appears only where we checked recent posts. Method: https://b2binfluencer.marketing/methodology/