Introducing VARA: The AI Business Process Consultant That Transforms How Enterprises Discover Automation

The intake interview is the most important conversation in the automation program. It determines what gets built, what gets recommended, and what gets the organization's development investment. It should be conducted with the care and rigor of a senior consultant - problem-first, data-driven, skeptical of easy answers, and comprehensive enough to produce a qualified recommendation.

The problem is that senior consultants are expensive, limited in availability, and inconsistent in approach. Even the best CoE teams apply their intake methodology differently depending on who conducts the interview, how much time they have, and how familiar they are with the specific domain of the process being discussed. The intake that determines a $500K automation investment is often conducted in a 90-minute workshop with a variable-quality spreadsheet and a BA who is splitting their time with three other projects.

VARA was built to fix this. Today we're sharing a detailed look at how VARA works, what it's built on, and what makes it different from the AI assistants and chatbots that have come before it.

What VARA Is Built On

VARA is a reasoning system grounded in the Problem-First qualification methodology - the four-gate framework that structures every discovery interview around problem definition, root cause analysis, pattern matching, and backlog refinement. This methodology isn't a script that VARA reads from; it's the logical architecture that VARA reasons within. VARA understands why the root cause question must precede the pattern question. It understands why asking 'could the system be fixed instead?' is more important than asking 'which automation platform would you use?' It understands why developer complexity matters for the backlog, not just the build.

VARA is not a fine-tuned domain model - it is a large language model (selectable between OpenAI, Anthropic, and Google Gemini) guided by a carefully designed system prompt that embeds the Problem-First framework, applies conversational best practices from enterprise consulting, and maintains the right balance between structure (following the four gates) and adaptability (adjusting questions based on what it learns).

How VARA Conducts an Interview

VARA opens every interview the same way: 'What specific process or workflow are you trying to improve, and what business problem is it causing today?' This is Gate 1 - problem definition. The question is designed to focus the business user on the problem, not the solution. From the first response, VARA begins adapting: probing for specifics if the answer is vague, asking about business impact if the problem is clear but the stakes aren't established, and moving forward when it has a clear, mature problem statement.

Gate 2 is VARA's most important and most distinctive moment. Before any discussion of automation, VARA asks directly: 'Before we explore automation options, I want to make sure we're solving the right problem. Could this issue be solved by fixing, upgrading, or replacing the underlying system rather than adding automation on top of it?' This question is asked of every process, every time. And VARA listens carefully to the answer - probing if it's unclear, following up if the answer implies a system limitation that the business user may not have fully considered.

The Deep Dive Capability

One of VARA's most valued capabilities in enterprise deployments is Deep Dive - the ability for CoE administrators to trigger targeted follow-up questions on specific intake responses they want to probe further. When a reviewer reads an intake report and sees that the business user gave a vague answer to the exception handling question, they can annotate that response with guidance for VARA - 'ask specifically about what happens to month-end close exceptions' - and VARA crafts a professional, context-appropriate follow-up question that appears in the business user's interview as a natural continuation of the conversation.

The business user never knows a human reviewer was involved. From their perspective, VARA is simply asking thorough, relevant questions. From the CoE's perspective, they've surfaced specific information gaps without scheduling another workshop.

What VARA Produces

• A structured qualification report: problem summary, root cause finding, recommended automation pattern with rationale, qualification path (showing which gates were passed and how), developer complexity recommendation, and ROI estimate.
• Structured intake data: all fields captured in structured format - volume, handling time, error rates, system information, document characteristics, compliance requirements - ready for use in prioritization and development briefing.
• A Mermaid.js process map: AI-generated current-state and future-state process diagrams based on the intake conversation.
• A scoring profile: five-dimension scoring (speed to value, TCO, complexity, risk, business impact) with an aggregate priority score.

Provider Flexibility and Graceful Fallback

VARA is AI-provider-agnostic. Organizations can configure IntakeOS to use OpenAI (GPT-4o), Anthropic (Claude), or Google (Gemini) as the underlying model, depending on their organizational preferences, existing enterprise agreements, or data sovereignty requirements. When no AI provider is configured, VARA falls back to a rule-based question engine that still applies the Problem-First framework - slower and less adaptive, but fully functional. Enterprise deployments never experience a hard failure due to AI provider availability.