Architect — Analyzing Your Processes

Architect analyzes your documented processes together to give you the big picture: what exists today, what's missing, what can be improved, and what to do next. It reads the workflows documented in Scout, the procedure documents you've uploaded, and the processes imported into the Processes library from your existing documentation — so you can run an analysis even for an Initiative with documents and imports but no interviews yet.

A process on the map keeps the name it has in the Processes library. A finished interview is filed into the library before any analysis runs, and Architect documents it under that same name rather than inventing one — so the row on the map and the process in the library are one process, not two. Two library processes that share recordings of the same activity (for example, confirmed variants) stay two rows on the map.

A process Architect knows only from documentation — no interview or recording behind it — carries a Documented, not observed badge in the process table, in its detail view, and on any gap raised against it. Its gaps and opportunities describe what the documentation says happens, not what anyone was seen doing; record a Scout session for it to confirm. The detail view lists the library processes it was built from under Documented in the Processes library, with a link to each. When the document a library process was imported from is also in an analysed collection, Architect reads the library process and leaves the file itself out, so the same procedure is not analysed twice and does not appear on the map or in the Library a second time. To analyse interviews and uploaded documents only, turn off Include imported library processes under Settings → Project brief.

Architect also reads the Initiative's written documents — facts pasted into the Library, notes saved from chat, answers recorded to its own questions, and short .txt / .md documents kept in Library folders — as clearly-labeled background on every run: useful, but weighed below what interviews show. They live in one place — the Library, under Documents — where they are read, added, edited and deleted; the Run Analysis window counts them and has an Open Library button. Adding one counts as new input for the next run. The Learned from interviews document Scout keeps there is read too — Architect uses its terms, system names and rules when it reads the interviews — but since it restates the interviews, updating it doesn't by itself make the analysis out of date.

Running an analysis

Only organization admins and admins or owners of the selected Initiative can start or re-run an analysis. Other members can open Architect and read the results, but they will not see Run Analysis, Run First Analysis, or Re-run architect.

1. Go to Architect from the sidebar.
2. Click "Analyze" (or "Re-analyze"). When re-analyzing, you can choose to build on your previous analysis so only new or changed interviews are re-processed.
3. Optionally add instructions for this run ("focus on compliance gaps", "prioritize automation"), then start it.
4. A progress popup shows each stage. A full analysis usually takes around 30–40 minutes — up to about an hour and a half with many departments and recordings. You can leave the page; the run continues in the background.

By default Architect analyzes the selected Initiative. Org admins can switch the Workspace dropdown to "All Initiatives" and run it across the entire organization, surfacing opportunities that span departments.

Architect is organized into As-Is views (what exists today — visible to everyone) and To-Be views (planning improvements — admins and members with To-Be access; see the Opportunities & Roadmap section).

When an analysis covers more than one department, a Department menu on the tab bar narrows every tab to one department. This does not run a separate analysis. A finding that also involves another department stays visible — an opportunity tagged to Accounts Payable still shows under Procurement when one of its processes is a Procurement process. Choose All departments to see the full run again. Processes that have no department appear only in that full view.

Overview

Departments and processes, how they connect, and which are documented vs. gaps. Admins click Run Analysis to populate it.

• Finishing new interviews never changes the Overview by itself. A notice appears saying the analysis is out of date, with a What changed line you can expand. Admins also get Re-run architect — until they re-run, what you see stays exactly as it was.
• A Preliminary badge means the results were drafted before your first full analysis — real and worth reading, but the Analysis chat, the open questions list, and challenging a finding all need a completed run.
• Click a gap status on a process row to open its details: description, suggested action, related questions/conflicts, and opportunities that address it. The actions sit in the pop-up's header: org admins see one Add to next round button (queues that gap for the interview planner), and everything else is in the ⋯ menu beside it. That menu holds Explain this gap (opens a chat about the finding) and, after a completed analysis, Challenge this finding; for org admins also Add all gaps to next round (queues every open gap in this analysis that isn't already in a round; if some already are, the item names how many are left) and Dismiss gap (out of scope — the gap disappears from the Overview and planner, and stays hidden across re-runs; the confirmation has an Undo); plus Redesign with AI (admins), View BPMN where a diagram exists, and Assign to department…. Once a gap is queued, the button is replaced by an In the next round badge and the menu offers Remove from round instead, until the round is sent.
• When the analysis is out of date, admins see an amber banner offering "Re-derive (fast)" — refreshes in minutes by reusing unchanged parts and folding in what changed — or "Re-run architect", a full rebuild. Prefer Re-derive unless something has gone wrong with the analysis itself. Other members still see that the analysis is out of date, without those buttons.
• The banner lists why it's out of date. Accepted corrections are only listed there when they change what a finding actually says — a gap or opportunity added or removed, a changed number, type, or recommended action, a correction to an interview SOP, or a change to your Initiative's assumptions. A typo fix or a reworded description doesn't bring the banner up, so when it does appear, re-deriving is worth it.

Fit/Gap

Compare your documented processes against vendor documentation (Oracle, SAP, or any system you're evaluating):

1. Upload reference documents — PDF, PowerPoint (.pptx), CSV, or Excel files describing the vendor system's capabilities; each is indexed automatically (a deck is indexed slide by slide). They are documents of the Initiative: the same files sit at the top of the Library under Documents, in Fit/Gap Reference Docs, and the section links there.
2. Run Analysis — the AI compares every documented process against your reference docs. Admins can steer every run with custom instructions (vendor context, target modules, system constraints) via the Initiative Settings button next to the heading.
3. View results — each process is classified Fit (green, fully supported), Partial Fit (amber), or Gap (red). Click a process to expand vendor module mappings, gap details, sub-step analysis, recommendations, and evidence. Summary cards show totals.

The tab only shows when Fit/Gap analysis is selected under Initiative Settings → Brief → Expected outputs (or when no expected outputs are selected, which means "run everything").

Conformance

Review process-mining conformance when log data is available.

Benchmark

Compare each as-is process to a cited industry reference model — not to something observed in your interviews. Built-in references cover Internal Audit (IIA Global Internal Audit Standards), Source-to-Pay, Order-to-Cash, Record-to-Report, Hire-to-Retire, and Financial Planning & Analysis. You can also import your own methodology (for example a published process map) with Use as industry reference model on the import dialog.

Each activity is marked Aligned, Diverges, Missing, or Extra, with the reference step named and a citation. Where a Scout interviewer asked someone who runs the process about a step or control their account hadn't mentioned (see Scout Interview → Questions about what you didn't mention), their answer is quoted under the row it bears on with an Interview evidence badge — From interview "Sourcing decision — Brazil": Is there an approval before the contract is signed? — Yes, the regional director above 50k. That is observed evidence, and the comparison prefers it over the SOP's silence: a "no, we don't" can mark a reference step Missing even when the step list says nothing, and a "yes, the budget owner, in SAP" can mark it Aligned. The line under the summary counts how many interview answers were used. Pick a process from the list on the left to see its activities; the reference models actually used are listed as labels at the top, with (yours) on any you imported. A radar per department scores it on Aligned, Complete, Conforming and Tight, above a count of how many activities aligned, diverged, are missing or are extra. Open a process from Overview and, if it matched a reference, the same comparison appears in its detail pop-up.

The comparison is built when an analysis run finishes. If it isn't ready when you open the tab, Linc builds it then — it can take a minute on a large analysis. Refresh comparison rebuilds it against the current references, for example after importing your own reference model. Processes that don't resemble any reference are left out, and if none matched you'll see a note saying so.

Divergences can become Opportunities tagged industry reference (a book icon) so they stay distinguishable from evidence-derived gaps. Turn that off under Settings → Brief if you want the comparison without those Opportunities. Excel/CSV export of Opportunities includes Origin and Reference columns. A presentation-style business-case deck from this comparison is not generated yet.

Backlog

Everything the analysis still needs from your team, in one ranked list: conflicts to reconcile first, then open questions and documentation gaps, ordered by what the answer unlocks. A question about a listed gap sits nested under it. Each row is labeled Gap, Question, or Conflict (two interviews described the same process differently), shows its department, and carries a status pill — Open, In next round, Sent, Forwarded (already out with someone by email), or Escalated (sent back to you by the person who was asked).

Click a question — anywhere on the row's text — to open the full ask in a pop-up: the question itself, Why it matters, the working assumption the analysis is currently running on, the department, the role best placed to answer it, and its status. The row's ⋯ actions are in the pop-up too, so you can answer it, forward it, or include it in the next round without going back to the list. You can also open a question from the keyboard: move focus to the row and press Enter or Space.

Gap rows aren't clickable. A gap's details open through its View gap button, which shows the same process pop-up the Overview uses. A question that is about a documentation gap has that button too — on the row and inside the pop-up.

An Initiative can be analysed for the whole Initiative and separately for individual departments. The Backlog merges all of those: a question raised by a department's own analysis sits in the same ranked list as the Initiative-wide ones, rather than being invisible until you switch scope.

Opening a question from a Details link elsewhere in the app (for example the interview planner's Next round list, or a disagreement's "see the question" link) scrolls to that question, highlights it, and opens its details straight away. Such a link also carries the question's scope: if it points at another Initiative the page switches to it, and if the question came from a department's own analysis the list opens filtered to that department, with Showing <department> only and a Show all departments button above it. Switching Initiative clears that filter.

A toggle at the top switches between Open — the work still to do — and Resolved, a history of questions your team answered and gaps that were dismissed. On the Open list you can also narrow to Conflicts (two interviews described the same process differently) or Ask IT/Security (questions generated from opportunities you selected on the Opportunities tab — they're tagged Ask IT/Security and grouped by opportunity).

• Admins get "Forward to answer" on each question: enter the email of whoever knows, and they get a link to confirm, correct, or say they don't know — no Architect access needed, just a sign-in. They need to be signed in with that email address on an account in your organization; if they open the link from another organization's account, the page tells them to switch accounts or ask you to re-send. Their answer arrives by email. Tick several questions and Forward selected to send them together in one email — the recipient can answer some now and the rest later. Org admins also get Include in round — use that instead when the answer needs a real conversation.
• A forwarded request stays open until it's answered, withdrawn, or the link expires after 7 days — re-running the analysis doesn't cancel it, so a late answer still counts and is picked up by the next run. If you answer the question yourself in Architect instead, the outstanding link is withdrawn and the recipient sees "This request was withdrawn".

Moving a process to another department

If the analysis filed a process wrongly, open its ⋯ menu → "Assign to department…". When the Initiative has departments in scope, only those are offered. Your choice normally sticks across re-runs — Linc remembers that a person set it by hand. The few cases where it can't be made permanent (an identically-named process exists, it's an undocumented gap, or its recordings belong to another process) are explained in the confirmation as you save. If a saved assignment stops applying later, an amber marker next to the process explains why.

Splitting, moving, or unlinking recordings

Architect decides which recordings document one process, and it errs on the side of merging: nine buyers who each recorded requisition-to-PO become one process, and because Compare works on processes, that one process can't be compared against itself. You have the last word. Open the process from the Overview and, in its Sources list, tick the recordings you want to act on:

• Split & compare — creates a separate process per person ("Requisition to PO — Amy Chen", "Requisition to PO — Bob Diaz", …) holding just that person's recordings, marks them as a confirmed variant group, and opens the comparison straight away — no re-run, no waiting for variant detection. Two recordings by the same person go to one variant. If the selection would only produce one variant (one person's recordings, none left behind), the button reads Split into variant and creates the process without a comparison.
• Move to… — moves the recordings onto another existing process in the same Initiative (search by name).
• Unlink — takes the recordings off this process without choosing another home for them. A process left with no recording and no document stays until the next analysis, which leaves it out unless new evidence for it appears.

Your decision sticks across re-runs. The next analysis puts each recording back where you put it — a split process is re-created under the same name with the same recordings rather than folding back into the original — and the confirmed variant group keeps pointing at them. The actions are unavailable while viewing a historical snapshot.

A recording tagged Q&A only is a follow-up in which someone answered questions without recording a screen. It adds detail to the process it sits on; it isn't a separate way of doing the work, so it's rarely worth splitting into its own variant.

To stop the merging up front instead — for an Initiative whose whole point is to compare the same process across people, branches, or regions — set Keep recordings of the same process separate by in the Initiative brief (Settings → Initiative Brief) before running Architect. Name the axis ("person", "branch", "region"), and the analysis keeps recordings of one activity by different ones as separate processes, named "<Process> — <axis value>", ready to compare. Leave it empty for the default behaviour.

Working with several processes at once

Select rows with the checkboxes to act on them together:

• Combined SOP — with 2+ connected processes selected, generate one end-to-end SOP covering the whole chain.

• Compare (Variants view) — select 2–8 versions of a process for a side-by-side breakdown of how they differ. The comparison shows an alignment score (of the steps every version documents, the share performed the same way), a coverage range per version, and an "alignment at a glance" strip color-coding each step. When alignment is very low, a banner explains why — an under-documented version, regional variants safe to harmonize, or possibly different processes grouped by mistake. A confirmed variant group has a page of its own in the Processes library (its name or Compare on the Library row opens it), which shows this same comparison under tabs shaped like a process page — Overview, Map, Steps, Standard, Changes — with the variants listed down its right side; see Processes → The variant group page. The ad-hoc comparison here is for any selection, grouped or not.

  The comparison has four tabs. The alignment and coverage badges sit on the tab row, with the version legend and a collapsed AI summary just under it, so the comparison itself is on screen at once:

  • Matrix (the default) — the compact step-by-step table: one row per canonical step, one column per version, and the recommendation.
  • Variants — a scorecard per version (steps performed, how many automated, roles and systems touched from its Activities grid, and how many changes adopting the standard would cost it), then each version's process map side by side, or the same thing as an aligned step grid (Maps / Steps).
  • Standard — the harmonization board. An impact strip at the top shows the standard as decided: steps in it, how many divergent steps are decided, changes required across all versions, how many version-steps go manual → automated by adopting (and how many would regress the other way), local variations and deferred steps. Below it, the standard diagram and the step-by-step standard. Under every step where the versions diverge, a Decision needed panel lets you record what the standard is: Adopt one variant's way (pick whose — the most automated is marked), Keep the recommendation, Write the standard yourself, Local variation allowed (the step stays, how it is done may differ by region), or Defer. Each decision can carry why the variation exists — Required (regulatory), Justified (business reason) or Incidental (habit, legacy) — an owner, and a rationale. Decisions save as you make them, are kept when the comparison is regenerated, and the impact strip, the Changes tab and the AI draft all follow them. Workshop mode walks the divergent steps one at a time on a large layout, each version's way side by side with Adopt this way buttons, for deciding in a room.
  • Changes — pick a version: the standard diagram colored by what adopting it means for that version, and its to-do list below (Change, Add, Drop; plus Local variation and Deferred where decided).

  A step a version appears to skip is shown as missing — confirm? until someone who knows says which it is: click it (in the step grid, or the chip under the step on the Standard tab) and choose Confirmed: not performed (a real gap, which then counts in the alignment score) or Not captured yet (a documentation gap, left out of alignment) — or Capture this step in Scout to go record it.

• Redesign together — with 2+ rows selected, open the Designer on the whole selection as one end-to-end map. In the Variants view the button reads "Harmonize" (limited to 8 at a time).

• Generate standard process — from a comparison (or a variant group's ⋯ menu), reconcile the variants into one harmonized standard process in the Designer, which you can refine and export as BPMN.

Gap analysis between two workflows

"How do I compare two processes?" has two answers, depending on what the two are.

Two versions of the same process (two people, branches or regions doing the same work — the usual meaning of "gap analysis between two processes"): use Compare.

1. Go to Architect (the sidebar) and open the Overview — the process map. Next to Group by, pick Variant: the map regroups into versions of the same process. Tick the checkboxes of the 2–8 versions you want to compare, then click Compare in the bar that appears. (A group that is already confirmed can also be compared from the Compare button on its row in the Library — no analysis needs to have run for that.)
2. Read the comparison: the Matrix tab is the step-by-step table (one row per step, one column per version); Variants shows each version's scorecard and map side by side; Standard is where you decide what the standard should be at every step that differs; and Changes lists what one version would have to change to adopt it. See Working with several processes at once above for the details of each tab.
3. A step one version appears to skip is shown as missing — confirm? Click it and choose Confirmed: not performed (a real gap, counted in the alignment score) or Not captured yet (it happens, but nobody recorded it; left out of the score). Until someone decides, the gap is only a suspicion.

If the two versions were recorded but you cannot tick two rows — because Architect merged the recordings into one process — open that process from the Overview, go to its Sources list, tick the recordings, and click Split & compare: each person's recordings become their own process and the comparison opens straight away. If a recording landed on the wrong process, tick it there and use Move to… to put it on the right one first. (See Splitting, moving, or unlinking recordings.)

Two different processes (say, Purchase Requisition and Vendor Onboarding): there is no side-by-side A-versus-B comparison — Compare only lines up versions of the same work. What you can do instead:

• Read each process's own gap analysis: on the Architect Overview, click the gap status on its row for the gaps, suggested actions and related questions; or open the process in the Library, where its Analysis tab gathers what is still missing from its steps and the gap notes from the latest Architect run.
• Ask Chat (the sidebar) to contrast them — "What does Purchase Requisition do that Vendor Onboarding doesn't?" — and it answers from both processes' published SOPs, with citations you can open.