SourceLace Docs
FAQOpen the app

Business setup

See how ready your setup is, review suggested setup changes, define your business terms, and save or bring in your whole SourceLace setup as one file. Nothing changes until an admin approves it. Business setup is new, and we keep improving it with what customers tell us.

About your setup

Business setup โ†’ About your setup is the questionnaire "Tell us about your setup": your systems, teams, sign-in and goals. The same questions are on Getting started, so you can answer them from either place, and your answers are the same in both. They decide your Getting started steps (for example, each system you list gets its own Connect step) and help you judge what the setup assistant suggests. Admins answer and change them at any time; steps already done stay done. Never type a password or key into an answer. The Business setup overview starts with how many questions you have answered.

Readiness score

The Business setup page opens with your readiness score, from 0 to 100, and the checks behind it, grouped by area: sign-in, access, data protection, sources, AI model, agents, skills and business terms, and audit and retention. Each check says whether that part is ready, almost ready or not yet, and why. Checks give counts, such as how many people are in no group, and never name people.

  • The checks that matter most, such as single sign-on, access by group and data protection, count three times as much as minor ones. A check that is almost ready counts half.
  • Checks of things your plan does not include are shown as not on your plan and left out of the score.
  • Most checks offer a fix: the page where you make the change, a setup assistant scan of one of your sources, suggested changes for the review queue, or an industry pack that fits your sources. A fix never changes a setting by itself: you approve each change in the review queue.
  • You can snooze a check for 30 days. It moves to the end of the list and still counts in the score, so the score always shows your setup as it is.
  • Your SourceLace contact can see your score and which checks are ready, but not the details of each check, and never names or email addresses. Each time they look, it is recorded in your audit log.

The readiness score is on every plan.

Review queue

Some parts of SourceLace suggest setup changes instead of making them: a setup file you bring in, and later ready-made setups for your industry and suggestions from SourceLace itself. Each suggestion arrives as a group of changes on the Business setup page, under Review, with the reason for each and what it would change.

  • Changes are grouped by the system they are about: one fold per source, such as Salesforce or SAP, with how many changes are waiting, approved and rejected. The first fold with changes waiting is open and the rest are closed; select a fold's name to open or close it. Inside a fold, each scan or setup file is a section, with the date it was made and how long it is kept, and its changes are grouped by kind (groups, access, protection rules, business terms, agents). Changes that are not about one system, such as a setup file, are grouped by where they came from. Search by what a change does to find one quickly, and tick Show decided changes too to see how earlier ones ended. The Review tab shows how many changes are waiting.
  • Approve or reject each change, all changes of one kind at once, or Approve all for a source (offered while its fold is open; if it includes turning on access by group, you are asked first, because that can lock out people who are in no group). Only approved changes are made, and they are made the same way as on the other admin pages, so the same checks and limits apply.
  • Some changes depend on others, such as access for a group that does not exist yet. They are made in the right order, and rejecting a change also rejects the changes that need it.
  • A change that needs a password or key (such as a new source) asks you for it when you approve. SourceLace saves it in the setting itself, encrypted, and never shows it again.
  • Groups that come from your identity provider, and the people in any group, are never changed by a suggestion.
  • New agents start in shadow mode (or as drafts when shadow mode is off), owned by the admin who approved them, so you can try them before they change anything.

Suggestions you leave alone are deleted after 30 days by default (an admin can change this under Limits). Every decision is recorded in your audit log.

Setup assistant

After you connect a system, SourceLace's own AI can read its structure and suggest a setup for it: groups and what they may read, data protection rules for fields that look sensitive (such as national ids, bank details, birth dates and personal phone numbers), business terms such as ARR or open pipeline, and the agents that fit. Every suggestion goes to the review queue, with the reason for it, and you approve each one with one click. Nothing changes before that.

Open Business setup, Assistant, choose a source you have connected and what you want suggestions for, and start. It takes a minute or two; the page tells you when the suggestions are ready to review. The tabs read in the order of the work: Your answers, then the assistant (or an industry pack, or a setup file) to make suggestions, and Review to decide on them.

  • A scan reads a system with the sign-in you connected to it, and keeps only counts, never values. For a source with a personal sign-in that is you, with your own permissions in it. A source with a shared account (one login set by an admin for everyone) is read with that account's permissions instead, and the page says Connected (shared account). Before you start, the page lists your organization's sources and says for each whether it is connected or Connect first. A source you have not connected yourself cannot be scanned: connect it under My connections first, and the page says so under the button.

  • Run it with an admin account that can see everything in that system. The assistant reads the system with the sign-in you connected (your own account for a personal sign-in, the shared account for a source that has one), so it can only suggest rules for objects and fields that account can see.

  • Check sample values (on by default) lets the assistant look at a few rows of each object to spot values such as Social Security numbers or IBANs in fields whose names do not say so. It keeps only counts, such as "18 of 20 values look like a Social Security number". The values themselves are never saved, logged, shown to anyone or sent to the AI. Health records (FHIR) are never sampled. Your admin can lower how many rows are checked, or turn it off, in Limits.

  • The AI sees only names: objects, fields, their types, short lists of allowed values, and those counts. Each scan counts as one AI question.

  • Fields whose names or values look sensitive are always suggested, whatever the AI says. Groups get read access unless you ask otherwise, and new agents start in shadow mode (or as drafts) so they change nothing until someone makes them live.

  • Turning on access by group comes last, and says how many people are in no group yet and would lose access.

  • Each recent scan links to its suggestions. Under Recent scans, a scan names its source, opens its suggestions in the review queue, and says how many are still waiting, approved or rejected. In the queue, a suggestion that replaced an older one from an earlier scan says so.

  • A group, term, agent or "turn on access by group" you rejected does not keep coming back. When you scan again, the assistant leaves out what your organization rejected before and says how many of each kind it skipped (it also skips a change that needs something that was rejected, such as access for a group you rejected). A suggestion counts as the same one when it has the same kind and name; access for a group is also per system and object. Data protection rules are the exception: fields that look sensitive are always suggested again, because you may want to protect them after all. SourceLace remembers a rejection only while the batch it was in is kept (30 days by default, then it may suggest it again), and an item you approved later ends it (one that was already in place when you approved it does not, and a rejection returns if the approval's batch is deleted before the rejection's). To see skipped suggestions again right away, tick Include suggestions I rejected before you scan, or choose Scan again with them on a scan: it starts a new scan of that source with them included.

The setup assistant depends on your plan, and your organization can run a limited number of scans a day.

Business terms

Tell SourceLace what your words mean: what counts as ARR, what "open pipeline" is, which field holds a customer's region. Add a term with its other names and a definition, and optional hints for a particular source. Admins edit terms under Business setup, Business terms.

The AI uses your terms everywhere SourceLace answers: in the SourceLace app, in agents, and in your own AI app, where they appear as the skill "Business terms". People can read that skill but change the terms only on the Business terms page. Each term here is marked Organization.

A group can also give a term its own meaning, on its page under Groups (marked From Sales, for example). For that group's people, the group's meaning wins. See Group terms, starter questions and default sources.

Matching

Business setup โ†’ Matching holds your organization's matching rules: how SourceLace recognises the same customer, supplier or person in two systems, such as a Salesforce account and a NetSuite customer. Write a rule, test it on live data, then publish it; a published rule applies to every question and agent that combines the two objects. Draft from our systems has the assistant draft a rule for two objects you name, in a chat. People's suggested rules wait at the top of Business setup โ†’ Review. See Matching records across systems.

Setup files

Under Business setup, Setup files, an admin can download the organization's setup as one file: access groups and what they may read, data protection rules, business terms, skills, agent templates, agent settings, the AI model, single sign-on, data retention and the limits you set. Agents are included only when you ask. Use it to keep a record, to review a setup, or to copy it from a test organization to your real one.

A setup file never holds people, group members, passwords, keys, chats or any data from your systems. When you bring a file in, SourceLace shows what it holds and which of your sources fits each part, lets you choose, and then puts only what differs from your setup now in the review queue. Bringing in the same file twice suggests nothing the second time.

Industry packs

Under Business setup, Industry packs, start from a ready-made setup for a department instead of a blank page. Packs are grouped by industry, then by the department they are for, and you can search them all by name, department or topic, such as "invoices" or "QBR". Every industry holds the packs for departments that work the same way in any business. Each pack gives a department its access groups, data protection, the business terms it uses, skills (short how-tos the AI follows), and agents that take recurring work off people's plates: digests that replace a manual report, prep before meetings, lists of only what needs attention, and email drafts. Each skill and agent says which manual task it replaces.

Agents also work with the tools your people already use, when you connect them: they draft emails in Gmail or Outlook (never sent: you review and send), post short updates in a Slack or Teams channel after you approve each post, and read files such as contracts or budgets from Google Drive, SharePoint or OneDrive. If you do not connect one of these, skip it: the agents still run and give you a report instead.

Several systems of one kind. Some parts of a pack take more than one of your systems, such as a Salesforce org for each region in Sales and revenue. Those parts show a tick box for each of your systems of that kind, all ticked to start with: untick any the pack should leave alone. Group access, data protection rules and the hints in business terms are suggested for every system you tick. When you tick two or more, choose how the agents are set up:

  • One agent that covers all of them (the default). An agent whose template can read several systems, such as the pipeline review, reads all of them in one run and says which system each finding came from. An agent whose template reads only one system, such as the deal desk queue, uses the first system you ticked, and the page says so.
  • One agent for each system, named with the system's name, such as Weekly pipeline review (EMEA org). The name is set when the agent is made and does not change if you rename the system later; if two systems have the same name, the system's short name is added.

You can run the pack again later, for example after you connect another Salesforce org and tick it. SourceLace then suggests adding that system to the agents the pack made, as items in the review queue: each becomes a new version of the agent that also reads the new system, with its mode and its actions' approval settings unchanged, and an agent shared with others asks its people to accept the new version. An agent you made yourself, or changed to another template, is left alone.

Every industry

  • Finance and FP&A (SAP S/4HANA, SAP ECC, NetSuite, Oracle Fusion or Dynamics 365 Finance and Supply Chain): Finance and FP&A groups, supplier bank details protected, terms such as DSO, DPO, gross margin and OPEX, skills for budget vs actual, the month-end close checklist, cash and AR aging, and invoice exceptions with Procurement, and agents for a weekly AR aging digest, overdue invoice reminders, month-end close status and a monthly budget vs actual.
  • Procurement (the same ERPs): Procurement, Budget owners and Finance groups, supplier bank details protected, terms such as spend under management, off-contract spend and three-way match, skills for spend analysis, supplier risk and blocked invoices, and agents for the PO approval backlog, supplier contract renewals (reading notice periods from your contract files), late purchase orders and a monthly spend summary.
  • IT service management (ServiceNow or Jira): Service desk and IT managers groups, detection of Social Security and card numbers in tickets, terms such as SLA breach, MTTR and first contact resolution, skills for major incident summaries, ticket trends and new hire access with HR, and agents for an SLA breach watch, the first update on each major incident, a daily backlog digest, weekly ticket trends and onboarding tickets for new hires (which you approve).
  • HR on SuccessFactors or Workday: HR and People managers groups, national ids hidden and pay, birth dates and home addresses visible only to HR, HR terms such as headcount and attrition, skills for headcount reports, a manager's team overview, attrition and aging requisitions, and agents for onboarding, a monthly headcount summary, work anniversaries, a monthly attrition watch and open requisitions aging. Agents must be tried in shadow mode first, for 30 days, and a person approves their changes for 30 days after they go live.
  • Leadership across CRM, ERP and HR: Leadership and Business operations groups, the same definitions of bookings, margin, DSO and headcount the department packs use, skills for the weekly business review, board pack prep and why a KPI moved, and agents for a weekly business review, a daily KPI watch that only speaks up when a number moves, and a board pack draft that follows your last board pack.
  • Customer Success, Marketing operations and Sales and revenue are listed here too; see below.

Software and technology

  • Sales and revenue (Salesforce, HubSpot or Dynamics 365): Sales, RevOps and Leadership groups, contact and lead phone numbers and email addresses masked, birth dates hidden where the CRM holds them, revenue terms such as ARR, open pipeline and win rate, skills for forecast calls, win-loss reviews, account briefs, quota attainment and deal desk reviews, and agents for pipeline reviews, data hygiene, stale deals, renewals, weekly quota attainment and the deal desk queue.
  • Customer Success (Salesforce, HubSpot or Dynamics 365, with your service desk, Gong, and Gainsight or ChurnZero if you have them): Customer Success, CS leadership and Support groups, terms such as health score and net revenue retention, skills for account health, QBR prep, churn risk and the handoff from Sales, and agents for a weekly health digest, Sev 1 escalation drafts, renewals at risk, renewal reminders and onboarding check-ins.
  • Marketing operations (HubSpot or Salesforce): Marketing operations, Marketing and Sales development groups, contact details protected, terms such as MQL, SQL and speed to lead, skills for campaign performance, lead routing and the MQL to SQL funnel, and agents for weekly campaign performance, lead routing hygiene, a weekly funnel and follow-up nudges for MQLs nobody has contacted.

Manufacturing and supply chain

  • Supply chain on your ERP (SAP S/4HANA, SAP ECC, NetSuite, Oracle Fusion or Dynamics 365 Finance and Supply Chain): Procurement, Planning and Finance groups, supplier bank details and tax numbers protected on SAP S/4HANA (on NetSuite and Oracle Fusion you choose which fields to protect), terms such as open PO, OTIF and days of supply, skills for supplier reviews, expedite lists, demand and supply exceptions and supplier on-time delivery, and agents for late purchase orders, low stock, demand and supply exceptions, supplier on-time delivery, month-end close and overdue invoices, and a freight invoice review when Flexport is connected.
  • Procurement and Finance and FP&A are listed here too.

With a data warehouse. Four packs also have an agent that reads a table in your data warehouse (Snowflake, BigQuery, Databricks, Amazon Redshift or Microsoft Fabric, among others) next to your CRM or ERP. Choose a warehouse as the pack's optional data warehouse source and they are set up in shadow mode; without one, they are simply left out:

  • Sales and revenue: a monthly ARR bridge (opening, new, expansion, contraction, churn and closing recurring revenue from your warehouse's ARR table, checked against the CRM's bookings).
  • Customer Success: a usage drop watch, which lists accounts whose product usage fell over four weeks, renewals first, with the owner and renewal date from your CRM.
  • Finance and FP&A: an ERP and warehouse reconciliation, comparing invoiced revenue or open receivables by month between the ERP and your warehouse.
  • Supply chain on your ERP: a forecast vs actual demand report, comparing the demand forecast in your warehouse with actual orders in your ERP.

Each one asks for the name of the table (or searches for one and says which it used), reads only, and posts nothing without a person's approval.

With Flexport. The supply chain pack also has a freight invoice review, which needs Flexport as its optional freight source (see Logistics: Flexport). Each week it lists the freight invoices from the last few weeks that deserve a second look: past due and unpaid, possible duplicates, charge types new for a shipment, and rates that moved more than you allow. It compares Flexport's invoices with each other, not with a contract, and covers only the freight Flexport handles. Without Flexport it is left out.

Healthcare

  • Healthcare on FHIR: groups for care management, clinical operations and patient access, extra detection on top of the masking of patient identifiers (which stays on), hospital terms such as readmission, length of stay and coverage problem, skills for readmission reviews and discharge follow-up, and six read-only agents, which must be tried in shadow mode for 30 days first:

    • a daily admissions digest;
    • a care gap focus list: patients most overdue for a test, such as HbA1c for people with diabetes;
    • a weekly readmissions watch, and a readmission alert that checks each patient as they are admitted;
    • a discharge follow-up check: patients sent home with no follow-up booked;
    • eligibility before the visit: tomorrow's appointments whose patient has no insurance coverage active on the visit date, so patient access can call ahead. It reads the coverage your EHR holds; it does not say what a patient will owe.

    Your EHR team activates SourceLace's app and allows it to read appointments and coverage (what they do). Not every EHR offers the hospital-wide searches the agents need: on Oracle Health only the eligibility check is set up, one clinic at a time, and the pack's page and suggestions say which agents were left out and why (which agents work with your EHR). Use it with test data only until business associate agreements are in place; it does not make your setup HIPAA compliant, and a qualified person on your team decides what to do with every result.

  • Healthcare claims and revenue cycle: for billing, revenue cycle and patient access teams that reach payers through a clearinghouse (Stedi, Availity or Optum). Groups for revenue cycle and patient access, extra detection on top of the masking of patient details (which stays on), terms such as denial, contractual adjustment and patient responsibility, skills for checking a patient's coverage, looking up a claim's status and explaining a remittance, and three read-only agents:

    • a weekly denial work list: claims denied or reduced for reasons other than the contract, by payer and reason code, the largest amount at stake first, with an optional email draft to the reviewer;
    • a remittance reconciliation: last week's payments and claims that do not add up, with the difference;
    • a payer scorecard: payers ranked by the share of claims they denied, with their most common denial reasons.

    It needs a healthcare claims source; the three agents read remittance, which Stedi and Optum sources answer (Availity sources do not yet). As with the FHIR pack, agents run in shadow mode for 30 days first, only admins create them, and the pack is for test data only until business associate agreements are in place: it does not make a setup HIPAA compliant.

Some packs also suggest templates to start from, such as meeting prep or QBR prep. These are agents each person makes for themselves in about a minute, because they work from that person's own calendar: open the template from the pack and answer a few questions.

Choose which of your sources each part is for, and the pack's contents arrive in the review queue. As with any suggestion, nothing changes until you approve it, anything you already have is left out, and agents start in shadow mode (or as drafts) so you can try them first. Packs for teams that work together share the same group names and definitions, so a second pack mostly adds to the first; read each suggested change, such as who may create agents, before you approve it. Each pack covers most of a typical setup; you still add people to the groups and adjust the agents to your needs, such as the channel to post in.

Help from your SourceLace contact

Your SourceLace contact or partner can suggest setup changes for you, such as an industry pack or a setup they prepared in a test organization. Their suggestions wait in your review queue like any other, marked as from them. On the Business setup overview you can also let them approve the changes they suggested; every approval is then recorded in your audit log under their name, as allowed by you, and you can turn it off at any time. Passwords and keys never travel with a setup: whoever approves a change that needs one enters it again.

Learn from use

SourceLace can notice questions several people keep asking in the SourceLace app, such as "what is the open pipeline in EMEA", and suggest turning each into an organization skill, or into an agent when the question comes back every week. Learning from use is off until an admin turns it on under Business setup, Learn from use, and an admin can turn it off at any time, which deletes its suggestions.

  • What admins see: the words the questions have in common, how many people asked, how many questions, the days the pattern was first and last seen, and which source the answers came from. Never who asked, and never the questions themselves. When people rate answers, it also shows how many of the answers got a thumbs-down, and suggests those first.
  • Example questions: only if an admin also turns on Show examples. Then up to three questions are shown per suggestion, each from a different person, with email addresses, phone numbers, card numbers, IBANs, Social Security numbers and your own data protection patterns masked. Other details people typed, such as a customer's name, can still show, so turn examples on only if that suits your organization.
  • When something is suggested: only when at least 3 different people asked 5 or more similar questions in the last 30 days (your SourceLace contact can change these numbers, and never further back than you keep chats). A word becomes part of a suggestion only if at least that many people used it. Patterns already covered by one of your skills or agent templates are not suggested.
  • Drafting: when an admin asks for a draft, SourceLace asks your AI model once, with the pattern's words, the counts and your sources' names (and the masked examples only if they are shown). It counts as one AI question. The draft skill or agent template arrives in the review queue; nothing changes until an admin approves it.
  • Dismissing: a dismissed suggestion is not suggested again. SourceLace keeps only a scrambled fingerprint of it, not its words.
  • How long it is kept: suggestions are deleted when the chats they came from are deleted, under your chat retention.

Learning from use only reads saved chats in the SourceLace app, never questions asked from your own AI app, and stays inside your organization: nothing is shared with other organizations or sent to SourceLace. It depends on your plan.

Setup files depend on your plan; see plans. Ask support@sourcelace.com if you would like to try Business setup.