AI-led discovery for complex software implementations

Sequential executive conversations. Objectives often come out broadly stated; conflicting priorities across units surface later than anyone would like.

Arthur interviews executives and business leaders in parallel, briefed to press toward measurable targets and to flag where priorities conflict across units or regions.

Executive input in days rather than weeks. Misalignment surfaces where it’s cheap to resolve — in Discover, not Design.

Captured at summary level through the same executive conversations; day-to-day process detail stays in people’s heads.

Arthur runs current-state interviews directly with process owners — a core Varos use case. The knowledge base begins here.

Genuine process depth from day one, inherited by every later phase.

A spreadsheet or portal form is sent for completion; in practice it often comes back incomplete or not at all.

Arthur interviews the relevant contacts and produces the same intake — complete, validated, and cross-referenced.

A reliable intake with no chasing. Also the client’s first, low-stakes interaction with Arthur — building comfort ahead of deeper phases.

With no SOW to allocate against, SMEs are borrowed ad hoc from other commitments; scheduling rarely lines up with the pursuit timeline.

SME involvement shrinks to reviewing Arthur’s draft and joining one short session — hours, not a workshop commitment.

Removes a structural constraint budget can’t fix. Pursuits stop waiting on SME calendars.

Pre-SOW, the client typically grants around two hours; the fit-gap is done at whatever depth that window allows.

Arthur pre-interviews stakeholders per theme and drafts the fit-gap with evidence and open questions. Live time goes only to contested items.

The scarce client hours produce a substantive fit-gap → tighter SOW scope, narrower estimation ranges, fewer surprises baked in.

Multi-day group workshops per functional area. The person with the deepest detail may not speak up in a full room; assumed alignment goes untested until sprints.

The same session prep materials the facilitating BA works from become Arthur’s brief. He interviews each participant individually, in parallel — 15–30 minutes — with follow-up rounds, mapping aligned vs. open items. Foundational sessions (e.g. data model) still gate everything downstream.

Multi-day workshops become one to two days of parallel interviews. Every voice is heard; alignment is tested systematically rather than discovered in Sprint 4.

Decisions happen inside the large workshops, and some items get lost in the room. End-of-week playbacks and extended SME sessions validate the week’s work.

The single live meeting — ~1 hour, run directly off Arthur’s output: what’s aligned is framed, and the agenda is the open items he flagged. The transcript feeds back for consolidation. The same consolidated record serves the extended SME playback — the broader group validates a record instead of re-hearing the week.

Roughly two days of workshop time becomes about an hour, and that hour is purely decisions.

Whether stories come from an AI generator or a BA template, the input is assembled by hand from workshop notes over several hours, and items get missed.

Your story-generation tooling stays exactly where it is. Its input becomes Arthur’s consolidated, conflict-resolved output in the template it expects, plus the decision-session transcript.

Output quality follows input quality — this raises the ceiling. No manual note-assembly in between.

The BA schedules an elaboration session per story to land the granular requirements — exact rules, formats, edge cases. Ambiguity often only appears once developers start.

Before any session, Arthur interviews the right client SMEs story by story — pulling out the exact rules, thresholds, formats, and edge cases that never made it into a document. Sessions stay for complex stories, but the BA arrives with drafted requirements. Simple, uncontested stories flow straight down the toolchain without one.

Complex sessions start from a draft and finish faster; simple stories skip the session — a saving that repeats every sprint. Developers stop stalling on mid-sprint ambiguity.

When a scope question arises, someone reviews months of notes to establish whether an item was ever discussed — determining billable CR vs. missed scope.

Every requirement is timestamped and attributed in the knowledge base, alongside the change baseline. “Was this discussed?” becomes a direct query. The CR process itself is unchanged — people decide; Arthur supplies the evidence.

Evidence available to both sides in seconds. Legitimate CRs supported, disputed charges avoided — protecting revenue and the relationship.

Handover material is assembled from scratch at project end, and much of the project knowledge leaves with the implementation team.

Most of the content already sits in the knowledge base, captured continuously since Discover. Handover becomes packaging — training guides, runbooks, a queryable knowledge base — plus a short close-out interview.

Handover effort drops sharply; the client receives a queryable knowledge base rather than a static deck, and key-person risk falls on both sides.

Limited instrumentation for outcomes; satisfaction gathered informally, and the loop back into the next wave rarely closes.

Arthur interviews users on feedback, satisfaction, and realized value — mapped against the targets set in Discover. The findings seed the next wave’s Discover.

Voice-of-user at scale, and value claims backed by structured evidence — exactly what renewal and next-wave conversations need.