Arvind Murali

In a Smarchitect Shortcast conversation with Microsoft data-platform executive Karthik Ravindran, they make four points: AI without context is vanity; governance enables rather than slows AI; most AI never reaches production because teams either overbuild a perfect foundation or overtrust a prototype; and the realistic future is people working with AI rather than being replaced by it. At Informatica World, they return to the same idea: metadata becomes memory, governance becomes trust, and context becomes intelligence.

Arvind Murali points to Informatica's 2026 CDO Insights report, which they say found that 69% of companies have integrated generative AI and roughly half are moving toward agentic AI, while data reliability and governance remain the biggest barriers to scaling.

• They recommend starting with one valuable use case, building the data and context it requires, learning, and then scaling.
• They attribute the context and governance framing to their conversation with Karthik Ravindran.

Read original · Original 2

They map enterprise architecture into four connected layers: systems of engagement (human plus AI experience), systems of intelligence (agents and orchestration), systems of context (data, semantics, governance, memory), and systems of record (CRM, ERP, ITSM). They single out context: an intelligent agent without trusted enterprise context cannot identify the right customer, policy, approval, entitlement, or action.

Murali illustrates the shift with the sequence data to context to decisions to actions to outcomes, arguing that it separates generative AI from what they call the agentic enterprise. They cite Databricks research that multi-agent systems grew 327% in four months and link agent deployment to their firm's delivery model: forward-deployed engineering, daily sprints, and human-in-the-loop governance.

Read original · Original 2

If it does not own the workflow, they say, it has no business owning the outcome. Agentic AI, they argue, shifts the economics from hours and utilization to completed work and measurable impact.

Murali proposes a hybrid commercial model: time-and-materials or fixed fee for discovery, fixed fee plus an outcome component for implementation, and shared-risk, outcome-linked managed services in production.

• They call the target state Intelligence-as-a-Service and name the metrics they would align both sides on: cost-to-serve, revenue growth, SLA adherence, first-time-fix rates, customer retention, and cycle time.
• They would not sell outcomes for marketing engagements, where media spend, pricing, competition, and customer behavior are outside the consultancy's control, or for governance-heavy AI systems that are probabilistic rather than deterministic.
• Their rule: never price an outcome they cannot directly influence; price the workflow they can control.

The model Murali recommends is one they say their employer, iOPEX Technologies, is already building, making the argument a description of their firm's commercial approach.

Read original

They argue that attention should go to business processes, workflows, and knowledge rather than technology, and that lack of trust and adoption are the biggest reasons AI solutions fail to reach production.

At a Databricks user group, Murali describes their firm's DemandPilot revenue operations solution and says the hardest part was not assembling the technology but getting more than 50 marketing analysts to adopt and trust the agents.

• They say agent output must be tested over time with realistic questions—in their case, over a three-month window—and that measurement should combine query, latency, and token dashboards with business measures such as average order value and return on ad spend.
• At a manufacturing and AI session, they describe a workforce gap: Gen X workers have deep tacit knowledge, Millennials pick up AI quickly, and Gen Z arrives AI-fluent but inexperienced.
• They propose continuous enablement and call using AI to surface tacit knowledge, objections, and fears 'AI for AI.'

Read original · Original 2 · Original 3 · Original 4

Experienced technicians know when not to trust a system; newer workers are more comfortable with AI. They propose capturing what workers observed and decided, why they made those decisions, and what happened next, then making that experience usable as governed context.

Murali describes building a two-way apprenticeship between technological fluency and operational wisdom using copilots, digital twins, knowledge graphs, and human-in-the-loop workflows.

• They ask how organizations are preserving tacit knowledge.
• On security, they argue that long-standing data-team disciplines—data quality, lineage, governance, access policies, tagging, classification, and customer master data management—are becoming central to cybersecurity.
• Security, they add, is not a one-time activity because attackers have the same AI tools as defenders.

Read original · Original 2

They argue that standalone CDP and MDM products are becoming features inside the data platform, shifting the economics and architecture for buyers.

Murali cites Gartner's projection that 80% of new CDPs will be built into the data platform by 2030, and lists the consequences: simplified architecture, AI as the orchestration layer for stewardship and data quality management, embedded governance such as Unity Catalog, zero-copy federation to systems of record, a shift from user-based to compute-based pricing, and demand for continuous, always-on intelligence.

• In a separate post, they compare Databricks' Lakebase Postgres offering with Supabase, AWS Aurora Serverless, and PlanetScale on branching for schema changes, vector search, autoscaling, and real-time lakehouse sync, based on their own application building.
• They also promote their employer's ElevAIte and MigrAIte service operations and revenue operations offerings at Salesforce Agentforce and Dreamforce events.

Read original · Original 2 · Original 3 · Original 4

Murali argues that consultants should not price outcomes they cannot control.

• They distinguish marketing results, which depend on factors such as media spend and product pricing, from operational workflows a consulting firm can own.
• They propose a hybrid model: discovery for a time-and-materials or fixed fee, implementation for a fixed fee plus an outcome component, and production through shared-risk managed services.

For brands planning a campaign

A webinar could use this distinction to help buyers evaluate outcome-linked AI contracts. Ask participants to map who controls each workflow and metric before discussing pricing.

In a conversation with Microsoft data-platform leader Karthik Ravindran, Murali organizes enterprise AI readiness around context, governance, people, and purpose. They recommend starting with one valuable use case, building the needed data and context around it, then learning and scaling—rather than waiting for a perfect foundation or treating a prototype as production-ready.

For brands planning a campaign

This post connects data-platform strategy to adoption and production. A useful session could ask attendees to identify one use case and the context and governance it needs.

Murali lays out four connected enterprise layers: engagement, intelligence, context, and systems of record.

• They single out context—data, semantics, governance, and memory—as the layer that helps an agent identify the right customer, policy, approval, or action.
• The post frames the shift as moving from prompts to deployed objectives and from data to decisions, actions, and outcomes.

For brands planning a campaign

For a data-platform or enterprise-AI campaign, this offers a clear architecture-led conversation. Ask where the proposed agent gets trusted context and how its actions connect to existing systems of record.

Describing an iOPEX RevOps solution, Murali emphasizes implementation questions beyond the technology: earning the trust of more than 50 marketing analysts, benchmarking realistic agent outputs over time, and measuring both system traces and campaign outcomes such as average order value and return on ad spend. They say analysts test several campaigns before choosing one.

For brands planning a campaign

This makes a practical case for a webinar on evaluating AI in revenue operations. Invite buyers to discuss adoption, benchmark design, cost-benefit tradeoffs, and business metrics—not just agent features.

Murali frames manufacturing AI as a knowledge-transfer problem: experienced workers know when not to trust a system, while newer workers may be more comfortable with AI. They suggest recording what workers observed, decided, and learned, then making that experience usable through governed context, copilots, knowledge graphs, or human-in-the-loop workflows.

For brands planning a campaign

A manufacturing AI campaign could focus on preserving operational knowledge while keeping experienced workers involved. Use this post to prompt discussion of what knowledge to capture and how workers can validate it.

They name context as the critical layer between data and action, then lay out a chain they return to: data to context to decisions to actions to outcomes.

• In a Smarchitect Shortcast conversation with Microsoft data platforms executive Karthik Ravindran, they turn that into four points: context makes data useful; governance enables rather than slows AI; most AI never reaches production because teams either over-build a perfect foundation or over-trust a prototype; and the realistic future is people working with AI rather than being replaced by it.
• At Informatica World, they make the same case from a different angle, arguing that metadata becomes memory and governance becomes trust, and citing Informatica's 2026 CDO Insights report that most companies have integrated generative AI but data reliability and governance remain the biggest scaling barriers.

They build a causal chain or numbered framework rather than list product features, anchoring it to a named executive conversation or vendor research report.

For brands planning a campaign

Strong fit for a data management, governance, catalog, or MDM vendor that wants the “why enterprise AI stalls” framing rather than a product walkthrough. The four-layer and data-to-outcomes structures can anchor webinar sections, while the Microsoft executive interview format shows they can carry a partner-branded conversation.

Examples: Source 1 · Source 2 · Source 3 · Source 4 · Source 5

They argue that advantage is won or lost in the context layer, because an agent without trusted context cannot identify the right customer, policy, approval, or entitlement. They also report that multi-agent systems grew 327% in four months according to Databricks research, and connect agent deployment to their firm's delivery model of forward-deployed engineering, daily sprints, and human-in-the-loop governance.

Architecture-level explainer posts that draw on named platform-vendor research; they frame the shift with a contrast (“from generative AI to the agentic enterprise”) rather than a product pitch.

For brands planning a campaign

Suits an agent platform, orchestration, or CRM vendor looking for a conceptual explainer. The four-layer map can anchor a sponsored session, while the human-in-the-loop framing gives enterprise buyers a governance story rather than a demo.

Examples: Source 1 · Source 2 · Source 3

They argue that outcome pricing only works when the consultancy owns, develops, and operates the workflow; without ownership of the workflow, it has no business owning the outcome.

• They sketch a hybrid commercial model: time-and-materials or fixed fee for discovery, fixed fee plus outcome for implementation, and shared-risk outcome-linked managed services in production.
• They call the target state Intelligence-as-a-Service.
• They list the metrics they would align both sides on, including cost-to-serve, revenue growth, SLA adherence, first-time-fix rates, customer retention, and cycle time, and name two situations where they would not sell outcomes: marketing engagements, where many variables sit outside the consultancy's control, and governance-heavy AI systems that are probabilistic rather than deterministic.
• Their closing rule: never price an outcome you cannot directly influence; price the workflow you control.

A take-a-position post that states a recommendation, makes the case for it, then names explicit conditions where it does not apply.

The recommended model is one they say their own firm, iOPEX Technologies, is already building, so the post also describes their employer's commercial approach.

For brands planning a campaign

Directly aligned with a marketer's need to plan a webinar on outcome-based AI contracts. A data platform or AI services brand could build the session around the hybrid pricing model and two exclusion cases, with the pricing-metrics list as a takeaway asset.

Examples: Source 1

They describe presenting the DemandPilot revenue operations solution to a Databricks user group and answer the questions their team faced at large RevOps customers: the hardest part was not the technology but getting more than 50 marketing analysts to adopt and trust the agents; benchmarking agent output requires realistic questions run over time, in their case a three-month window; scale is real, with each analyst expected to create and manage a campaign a day; and measurement combines query, latency, and token dashboards with business measures such as average order value and return on ad spend.

• They frame their firm's focus as RevOps and ServiceOps transformation rather than chasing new tools.
• Their employer promotes ElevAIte and MigrAIte service operations and revenue operations offerings at Salesforce's Agentforce and Dreamforce events.

Implementation field notes with concrete operating numbers and named customer-facing problems; they answer audience questions directly and credit colleagues by name.

For brands planning a campaign

Fits a revenue operations, marketing operations, or service operations platform vendor looking for adoption and measurement detail rather than an AI theme. The adoption-over-technology finding and the AOV and ROAS measurement layer give a webinar a practical spine.

Examples: Source 1 · Source 2 · Source 3 · Source 4

They report building Lakebase-based applications and compare Databricks' Postgres offering with Supabase, AWS Aurora Serverless, and PlanetScale on branching for schema changes, vector search, autoscaling, and real-time lakehouse sync.

• In a separate post, they argue that standalone CDP and MDM products are becoming features inside the data platform, citing Gartner's projection that 80% of new CDPs will be built into the data platform by 2030.
• They list the resulting shifts: simplified architecture; AI as the orchestration layer for stewardship and data quality management; embedded governance such as Unity Catalog; zero-copy federation to systems of record; a move from user-based to compute-based pricing; and demand for continuous, always-on intelligence.
• At Informatica World, they point to MDM, catalogs, governance, metadata, and knowledge graphs as the foundations AI agents need for trusted identity, relationships, and context.

Comparison posts draw on their own build experience and include a diagram, alongside named analyst-firm projections supporting the architectural claim.

They frequently name Databricks, Informatica, Microsoft, Salesforce, and other vendor products. Their posts describe their own building and event activity; ask about any commercial relationships before planning a partnership.

For brands planning a campaign

Useful for a data platform or lakehouse-ecosystem vendor, especially a governance, catalog, or MDM product looking to position itself as context infrastructure rather than a standalone category. The compute-based-pricing observation offers a commercial angle for pricing or packaging discussions.

Examples: Source 1 · Source 2 · Source 3

From a manufacturing and AI session, they report that manufacturing is adopting AI from the shop floor to the contact center; lack of trust and lack of adoption are the two biggest reasons AI solutions do not reach production; and a significant gap exists between Gen X workers with deep tacit knowledge, Millennials picking up AI quickly, and Gen Z arriving with AI fluency but little experience.

• Their answer is continuous AI enablement and education, plus a practice they call AI for AI: using AI to surface tacit knowledge, objections, and fears so adoption can be smoothed.
• In a later manufacturing summit post, they frame the goal as capturing experience as governed enterprise context through copilots, digital twins, knowledge graphs, and human-in-the-loop workflows, creating a two-way apprenticeship between technological fluency and operational wisdom.

Conference field notes organized as numbered observations, followed by a question asking the audience how their organization handles the problem.

For brands planning a campaign

Good match for a manufacturing, industrial AI, or knowledge management vendor looking for a workforce-adoption narrative rather than a pure technology pitch. The tacit-knowledge capture angle can anchor a webinar on deploying AI alongside experienced staff.

Examples: Source 1 · Source 2

Disciplines data teams have followed for decades—including data quality, lineage, governance, role- and attribute-based access policies, tagging, classification, and customer master data management—are becoming central to cybersecurity.

• Their second point: security is not a one-time activity, because bad actors have access to the same AI tools as defenders.
• They also offer career advice to future leaders, recommending data analysis, analytics, storytelling, and articulating value propositions as skills for the age of AI.

Short event recap organized around two named learnings and a recommendation for a specific audience.

For brands planning a campaign

Suits a security or data governance vendor seeking a data-centric security angle, and shows they can address practitioners and early-career audiences in the same piece.

Examples: Source 1

They open with a provocative claim or question, then work through numbered points, often datelining the post to a conference, user group, or vendor event.

• They interview executives for their Smarchitect Shortcast and repackage a single long conversation into four named 'executive micro-masterclasses,' showing how they can turn one interview into several short-form teaching pieces.
• They cite third-party research from Gartner, Databricks, and Informatica, credit colleagues and partner-side people by name, and write informally, using caps and emoji for emphasis.
• They also mix career reflection into the same feed, as in a post about SpaceX employees' equity windfall that ends with their own employer and the vision they say they buy into.

Numbered-list and question-led writing, executive interview recaps, and architecture posts that pair a claim with a diagram or vendor comparison.

The posts are personal commentary rather than long-form analysis; several reference their employer's offerings, and one focuses primarily on career and employer vision rather than technology.

For brands planning a campaign

Planning notes for a brand: realistic formats include a sponsored explainer or numbered-lessons post, a conference recap around an event their team attends, and an interview-led shortcast series with executive guests. Their posts work best when a partner provides a specific claim, report, or product fact for them to frame, rather than a general topic.

Examples: Source 1 · Source 2 · Source 3 · Source 4 · Source 5 · Source 6 · Source 7 · Source 8 · Source 9

Posts use “we” and “our team” to describe iOPEX’s consulting model, DemandPilot, and event plans.

View example

Ask whether any commercial relationship with Databricks exists.

View example

Ask whether any commercial relationship with Microsoft exists.

View example

Ask whether any sponsorship or employment relationship with Informatica exists.

View example

Ask about any sponsorship or partnership terms.

View example

Ask whether the invitation involved a commercial engagement.

View example

Ask whether there is an ongoing relationship.

View example

Both writers treat trusted, well-governed data as the precondition for agents that act inside real workflows, and both cite the same vendor economics, including Salesforce's Informatica acquisition, to make that point.

Perla writes as a Salesforce and MuleSoft practice lead, aiming their posts at IT leaders working out what the platform is becoming and what replaces what in their stack.

• Murali ranges wider across data platform architecture, RevOps implementations, industrial adoption, and consulting pricing, reporting from their own building and conference activity rather than from one vendor ecosystem.
• Perla also closes some posts with a direct practice contact invitation, which changes the tone.

Read original · Original 2

Rice convenes senior data and AI leaders and writes about accountability, judgment, and who owns the outcome when an agent makes a decision. That aligns with Murali's position that whoever owns the workflow should own the outcome, and that governance enables rather than slows AI.

Rice's writing is leadership reflection built around executive convenings, with technology mostly as a backdrop and an emphasis on a leader's responsibility toward the people they are accountable for.

• Murali stays closer to architecture, adoption mechanics, and commercial structure.
• Rice also hosts and builds community around a named event series, giving readers a different role to follow.

Read original · Original 2

Bidgood's line that agents do not question context, they operationalize it, and that an incorrect ontology produces faster, confident answers that are still wrong, matches Murali's claim that the context layer is where agent advantage is won or lost. Both treat data foundation work as the real AI work.

Bidgood stays inside consumer goods, working through trade spend, retail media, product hierarchies, and supply chain alignment, so the examples are vertical and commercial rather than architectural. Murali moves between lakehouse and CDP architecture, pricing models, manufacturing knowledge transfer, and security, writing more often from events and implementations than from a single industry problem.

Read original · Original 2

Their posts take a clear line that Salesforce stopped being a CRM and the integration layer matters more than ever, written for enterprise architects and IT leaders governing a platform that keeps changing. For a Salesforce-focused webinar or sponsored piece aimed at that reader, they are closer to the decision than a cross-platform data voice.

Murali's frame is Salesforce-centric and reads as practice marketing.

• They run a MuleSoft and Salesforce practice, and one post closes with a direct approach for work, so the session risks becoming a capability pitch rather than an independent framework.
• A vendor wanting neutral comparison across platforms, or data-platform positioning beyond the Salesforce stack, gets less here.

Examples: Source 1 · Source 2

She convenes senior data and AI leaders, with a Dallas session at Microsoft's innovation hub and named featured executives.

• Her writing covers leadership responsibility, trust, and what organizations owe their people as AI changes work.
• For a campaign whose success is senior attendance, peer conversation, and a respected host, she brings convening and moderation value that a framework-first speaker does not.

The tradeoff is specificity.

• Her posts discuss leadership questions and quote others on where to place bets, but do not set out an operating model for outcome-based contracts or platform architecture.
• Ask whether she can provide a testable commercial framework; otherwise, the buyer would need to supply that content or add a second speaker.
• The commercial terms tie to her own event brand.

Examples: Source 1 · Source 2

Murali's posts make that case tightly and share the same underlying point as Arvind's: agents operationalize whatever context they are given. A CPG-focused session lands with an audience that recognizes its own problems.

The tradeoff is range.

• Both posts stay inside consumer goods, so a horizontal enterprise data-platform or services campaign gets a narrower conversation than from a creator who also covers platform architecture, agent governance, and commercial models.
• Because the core context argument overlaps with Arvind's, choosing Mark is mainly a vertical-focus decision rather than a different intellectual angle.
• The buyer should be clear which one they need.

Examples: Source 1 · Source 2