Apratim Purakayastha

The clean corporate story is that Vector Capital hired a veteran software executive to merge Showpad and Bigtincan and lead an “AI-native revenue effectiveness” company. That is true, but it misses why Apratim Purakayastha is unusually matched to the assignment.

Across three decades he has kept returning to the same systems problem at larger scales: how do you preserve local variation without paying the reliability, security, and operating cost of unmanaged fragmentation? His doctoral work characterized I/O behavior in parallel scientific systems before optimization; IBM put him on mobile, pervasive computing, WebSphere, Lotus, and the company’s first commercial SaaS effort; ACI and SevOne moved him from architecture into regulated on-demand P&Ls; SumTotal and Skillsoft added retention, platform consolidation, and full-unit economics; Showpad gives him the problem in its purest commercial form.

The continuity is an editorial inference, not a claim he has made about himself. But his own May 2026 formulation supplies the name: the opportunity for enterprise SaaS is not unrestricted customization. It is “manageable mass customization at scale.”

1. 1
   Find the costly fragmentationLegacy products, disconnected tools, inconsistent content, bespoke scripts, or divergent teams.
2. 2
   Define the common substrateShared platform, data, permissions, process, and measurement — without pretending every customer is identical.
3. 3
   Fit process to the systemRegulated payments, fast-release SaaS, learning platforms, and field selling need different release and control rhythms.
4. 4
   Learn with representative customersDesign partners across archetypes; do not generalize from ten similar customers or the loudest internal team.
5. 5
   Change behavior inside the workflowEducation, coaching, product, and services converge around the job to be done.
6. 6
   Measure the outcome, reopen the loopRetention, contribution margin, execution consistency, win rate, customer value — not feature count or course completion.

Purakayastha’s most revealing AI argument begins before generative AI. Lotus Domino agents and VBScript let enterprises encode millions of bespoke workflows. They also created abandoned scripts, broken links, stale schemas, and institutional knowledge nobody could safely maintain. SaaS won by trading some local flexibility for predictable cost, reliability, security, and scale.

Generative agents reopen the customization frontier. Natural language lets any employee create a micro-workflow against enterprise data. But without lifecycle controls, the enterprise simply recreates the Domino/VBScript estate with stochastic code. His answer is not “agents replace SaaS.” It is that SaaS becomes the governed substrate for agents: permissions, approved data, connectors, analytics, versioning, and common workflows below; context-specific execution above.

Showpad’s Summer ’26 release is the first visible implementation of that doctrine. Six deployable Genie Agents sit inside an Agent Studio; a remote MCP server lets third-party agents reach governed Showpad content and workflows; Salesforce, Microsoft Copilot, and Glean integrations keep agents inside enterprise permissions; offline field capture feeds structured context back into CRM; and analytics connect usage to revenue metrics.

In a 2023 interview, Purakayastha reduced enterprise AI adoption to three problems. First, people need role-specific awareness and hands-on exposure. Second, the organization must change processes, operating models, workflows, and controls. Third, leadership must mobilize the new work inside existing roadmaps rather than build a disconnected innovation theater.

That framework is stricter than the usual AI-enablement pitch. It treats learning as a precursor to changed behavior, not the deliverable. Purakayastha has also argued that passive instruction is insufficient: practice, mentoring, and realistic situations make learning tangible. His safety posture is similarly operational — humans need an escalation path; private data, bias, and ethical use have to be designed in before broad adoption.

This makes an advisory practice the natural execution layer between product capability and customer outcome: diagnose maturity, select role-specific workflows, establish controls, run design partnerships, instrument behavior change, and prove business impact. The advisor is not a trainer bolted onto the software. The advisor is the person who converts the software into a repeatable customer operating system.

Purakayastha’s March 2026 essay on B2B marketing is the bridge between his product history and a modern CMO remit. He rejects both the old “marketing tells the story” model and the self-contained pipeline-machine model. Buyers now validate claims through product experience, peers, customers, analysts, communities, and AI systems. Pipeline is therefore a delayed, company-level expression of product-market fit and accumulated trust — not a departmental output.

His useful phrase is expectation debt: features positioned ahead of readiness, outcomes implied but not delivered, narratives that are aspirationally true but experientially false. The debt shows up later as slower deals, weaker champions, discounts, churn, and fragile expansion. Marketing’s highest-leverage job becomes reputation alignment — pull product and customer truth into the narrative, then multiply corroborated proof across the surfaces buyers trust.

This doctrine also explains his insistence that product leaders sell consultatively while a new offer is still earning repeatability. Design partners, early adopters, customer archetypes, support signals, seller feedback, and workarounds form a learning loop. Only after the product has won enough difficult cases can sales scale the motion confidently.

Purakayastha’s essays on process, data misuse, and executive reality distortion supply the governance layer. Process is not inherently bureaucratic; the wrong process imposed on the wrong product is. Data is not truth when it is selectively sampled, averaged across incompatible segments, or used to defer a decision. Executive abstraction is useful only when it can be unfolded back into a faithful operating reality.

Across the record, his operating loop reads the same at every scale he has worked at: find the costly fragmentation, define the common substrate, fit process to the system’s risk and release physics, learn with representative design partners, change behavior inside the workflow, then measure the business outcome and reopen the loop. The measures that count are retention, contribution margin, execution consistency, win rate, and customer value — not feature count or course completion.

The candor requirement is not a soft skill in this framing. It is a structural property: if the executive summary cannot be unfolded into facts across representative customer segments, the loop is running on distorted signal and the substrate will be built for a customer who does not exist.

Company claims are labeled as claims; LinkedIn career metrics are treated as self-reported unless reconciled to filings. Research snapshot: 2 September 2026.

The thesis in §§ 01–06 is the strongest evidence-backed reading of the record. It is not the only reading, and four seams remain open. Each is a place where a defensible narrative outruns what public sources can currently prove.