03 · Operations
The Deal Workflow
End-to-end lifecycle of a mortgage at a non-bank CRE lender like Cameron Stephens. Use this as the map for where AI and tech can reduce cycle time, decision risk, and headcount load.
Executive framing
A CRE lender's value chain has thirteen distinct stages. The first six are the origination funnel (most deals die here; everything is judgment-heavy). The next four are the execution gauntlet (legal + ops dominated; checklist-heavy). The last three are the portfolio operating system (data-intensive, cyclical). AI and tech opportunities cluster differently at each phase — confusing them is the most common transformation mistake.
1 · Sourcing
| Who | BD / originator / managing director ("relationship runner"); occasionally direct from sponsor |
|---|---|
| Inputs | Inbound from brokers (JLL, CBRE, Colliers, boutiques), repeat sponsor relationships, internal referrals, conference / industry events |
| Outputs | Deal opportunity logged in pipeline (often spreadsheet or CRM); preliminary read on sponsor, asset, ticket size |
| Decision | Soft "look further" vs polite pass |
| Cycle time | Hours to days |
| Tech touchpoints | Email, broker portals, CRM (Salesforce, HubSpot, Dealpath), deal log spreadsheets |
Operator The biggest organisational sin at this stage is invisible: deals that pass through and are not logged. You can't manage pipeline you can't see.
2 · Screen
| Who | Originator + senior underwriter |
|---|---|
| Inputs | Sponsor's offering memorandum / package, pro forma, project plan, sponsor financials, prior project references |
| Outputs | Screening note: deal economics rough-cut, fit with current mandate, indicative pricing, gating risks |
| Decision | Issue term sheet vs decline |
| Cycle time | 2–10 days |
| Tech touchpoints | Pro forma spreadsheet (Excel), market data (Altus, Costar), CRM stage advance |
3 · Term sheet
| Who | Originator drafts; managing partner / Head of Lending approves |
|---|---|
| Inputs | Screening output |
| Outputs | Non-binding term sheet: loan amount, rate, fees, term, security, key covenants, conditions to commitment, exclusivity period |
| Decision | Sponsor signs / countersigns; pays good-faith / engagement fee |
| Cycle time | 1–3 days to draft; days to weeks for sponsor signature (they often shop alternatives) |
| Tech touchpoints | Word templates, e-sign (DocuSign), deposit handling |
4 · Underwriting
| Who | Underwriting team (analyst + senior underwriter); coordinated by deal lead. External: appraiser, QS, environmental, legal |
|---|---|
| Inputs | Full sponsor package (audited financials, project plan, budget, schedule, plans, pre-sale data, leasing data, market study, environmental Phase I, geotech, insurance evidence, title commitment, corporate structure, beneficial ownership) |
| Outputs | Underwriting model (Excel), market analysis, sponsor analysis, third-party reports collated, stress tests, risk grading |
| Decision | Internal: progress to Credit Committee, restructure, or decline |
| Cycle time | 2–6 weeks (longer if appraisal / environmental delayed) |
| Tech touchpoints | Excel models, project file (SharePoint, Box, Egnyte), valuation reports (PDF), market data subscriptions, KYC vendors |
Operator The third-party report cycle (appraiser, QS, environmental) is the most common cause of underwriting delay. Each is a PDF returned by an external firm; the underwriter parses, validates, and reconciles by hand. This is rich territory for document AI.
5 · Credit memo
| Who | Underwriter (draft) + Head of Lending (review) |
|---|---|
| Inputs | Underwriting model, third-party reports, sponsor & sponsor-team profiles, market data, comparable transactions, internal precedent |
| Outputs | Credit memo: deal summary, sources & uses, structure, sponsor, asset, market, exit plan, sensitivity, risks & mitigants, recommendation, conditions of approval |
| Decision | Memo signed off for Credit Committee tabling |
| Cycle time | 3–10 days |
| Tech touchpoints | Word template (highly variable across lenders), deal data room |
6 · Credit Committee
| Who | CEO/President, CIO/Chief Credit Officer, Heads of Lending, sometimes independent member; deal team presents |
|---|---|
| Inputs | Credit memo (circulated 2–5 days ahead), Q&A |
| Outputs | Minuted decision: approve / approve with conditions / defer for more work / decline |
| Decision | The binding internal yes/no |
| Cadence | Weekly or bi-weekly, plus ad-hoc for urgent |
| Tech touchpoints | Boardroom + circulation; minutes captured (often Word + email); historic decisions archived (often not searchably) |
Watchpoint
Credit Committee discussion is the single most decision-relevant data the firm produces — and at most lenders it is barely captured. If CS doesn't have searchable, structured CC minutes today, that's both an audit-risk gap and one of the highest-ROI AI opportunities (transcription + summarisation + linkage to deal records).
7 · Commitment & conditions precedent
| Who | Originator (relationship), legal (drafting), underwriter (conditions tracker) |
|---|---|
| Inputs | CC-approved terms, sponsor sign-off on commitment |
| Outputs | Commitment letter; conditions list tracked to satisfaction (typically 20–60 items: insurance evidence, KYC complete, title clear, opinion letters, etc.) |
| Decision | Each condition cleared or waived; readiness to close |
| Cycle time | 2–8 weeks (sponsor-dependent) |
| Tech touchpoints | Conditions checklist (Excel, Smartsheet, Dealpath), shared data room, e-sign, legal practice mgmt |
8 · Closing & funding
| Who | Closing team / loan ops, external legal counsel (both sides) |
|---|---|
| Inputs | Signed loan agreement, security documents, title commitment, insurance certificates, satisfaction of all CPs, payment instructions |
| Outputs | Charge registered (provincial Land Titles), wire issued, loan booked into servicing system, closing binder filed |
| Decision | Funded |
| Cycle time | 1–5 days (after all CPs satisfied) |
| Tech touchpoints | Loan servicing system (book the loan), wire system / bank portal, Land Titles (Teranet ON, LTSA BC, ALRO AB), document mgmt |
9 · Servicing & draws
Now the loan exists. Two operating modes:
Term loan servicing
- Interest invoiced monthly; payment received
- Property tax / insurance verified annually
- Annual borrower financials / rent rolls / op statements collected and analysed
- Covenant tests (DSCR, debt yield) run at defined intervals
Construction loan draws
- Monthly: sponsor submits draw request with invoices, lien waivers, updated cost-to-complete
- QS issues monitor report certifying work in place & remaining cost-to-complete
- Loan admin verifies, calculates advance net of holdback, releases funds, updates loan balance
- Statutory holdback retained until lien period expires post-substantial-completion
Operator Draw processing is mostly checklist work but with real money on the line. A typical construction loan has 15–25 draws over its life. The QS report is the gating document; getting it parsed and reconciled against budget faster is a quantifiable productivity win.
10 · Monitoring & watchlist
| Who | Portfolio management / asset management / credit risk team |
|---|---|
| Inputs | Servicing data, borrower financials, market data, news on sponsor or asset, macro / sector reports |
| Outputs | Quarterly risk ratings, watchlist, action plans for deteriorating loans, ECL inputs for accounting |
| Decision | Re-rate, watchlist, formal default notice, restructure |
| Cadence | Continuous; formal quarterly review |
| Tech touchpoints | Portfolio management system, BI dashboards, IFRS9 ECL model, news monitoring |
11 · Reporting
Three audiences, three different reports:
| Audience | What they want | Cadence |
|---|---|---|
| Institutional LPs (pensions, life cos) | Portfolio composition, performance vs benchmark, individual asset notes, ESG metrics, risk & concentration, NAV, capital activity | Quarterly; ad-hoc on incidents |
| Private investors (HNW, accredited) | Performance, distribution, fund-level commentary, statements, tax slips (T5008/T5/T3 as applicable) | Monthly statements; quarterly commentary; annual tax |
| Internal & regulatory | Board pack, IFRS9 financials, regulator/insurer filings, AML reporting | Monthly / quarterly / annual |
12 · Payout & discharge
| Who | Loan ops + legal |
|---|---|
| Inputs | Payout request, payoff statement, wire received |
| Outputs | Security discharged (Land Titles), PPSA discharges filed, closing letter to borrower, file archived, capital returned to investors / redeployed |
| Cycle time | 3–10 days |
| Tech touchpoints | Servicing system, Land Titles registries |
13 · Workout track (parallel)
When a loan defaults or is heading there, a different operating mode kicks in:
- Notice / forbearance. Default notice issued; often initial forbearance agreement allows breathing room with milestones & fees.
- Strategy meeting. Lender chooses path: continue, restructure, replace sponsor, sell loan, enforce.
- Enforcement. Power of sale (ON), judicial sale (other provinces), or court-appointed receiver. Specialist legal counsel.
- Take-back. Lender may end up owning the project — runs through Equity Capital side or external operator until disposed.
- Disposition. Sale of property or completed project; recovery proceeds applied.
- Loss recognition. Final write-off through ECL; reported to investors and regulator.
Watchpoint Workout is brutal on data — files were not designed for it, third parties (receivers, counsel) generate new artifacts, and the deal team morphs. A unified deal record from origination is hugely valuable in workout. Inverse: a chaotic origination record is a workout nightmare.
Systems map (typical Canadian non-bank lender)
| Layer | Typical tools | Stage(s) |
|---|---|---|
| CRM / pipeline | Salesforce, HubSpot, Dealpath (CRE-specific), Excel | 1–3 |
| Document repo | SharePoint, Box, Egnyte, internal NAS | 1–13 |
| Underwriting models | Excel (always) | 4–5 |
| Market data | Altus InSite, Costar, Yardi Matrix, Real Capital Analytics / MSCI, Realtors' boards | 2, 4, 10 |
| KYC / AML | Trulioo, Persona, LexisNexis, manual + FINTRAC reporting | 4, 8 |
| E-sign | DocuSign, Adobe Sign | 3, 7, 8 |
| Loan servicing / mortgage admin | McCracken Strategy CS, FIS Commercial Servicing, Sagen platforms, NetSol, custom — Canadian market is fragmented and many lenders run on in-house systems | 8–12 |
| Land Titles | Teranet (ON), LTSA (BC), SPIN/ALRO (AB) — accessed by legal counsel | 8, 12 |
| Portfolio / risk | Custom dashboards on top of servicing data; sometimes Argus, Yardi Investment Mgmt | 10 |
| Investor reporting | Juniper Square (US-built but used in Canada), Allvue, in-house | 11 |
| Accounting / IFRS9 | Yardi Voyager, Sage Intacct, Microsoft Dynamics; ECL modelling in Excel or Moody's RiskCalc / SAS | 10, 11 |
| BI / analytics | Power BI, Tableau, Looker on top of warehouse | 10, 11 |
Common bottlenecks (where AI / tech wins are visible)
| Bottleneck | Where | Symptom | Likely AI lever |
|---|---|---|---|
| Third-party report intake | Stage 4 | Underwriters re-keying PDFs into Excel | Document extraction → structured data into model |
| Sponsor financial diligence | Stage 4 | 2 weeks to parse a sponsor's tax returns + project histories | LLM-assisted financial statement extraction; entity resolution across past projects |
| Credit memo drafting | Stage 5 | Senior underwriter writing for days | Memo first-draft generation from underwriting model + diligence notes |
| Conditions tracker | Stage 7 | Email tennis with sponsor counsel | Workflow tool + AI-assisted condition status interpretation from incoming docs |
| Draw processing | Stage 9 | QS report → Excel reconciliation → wire | QS report parsing + reconciliation + audit trail; exception-only review |
| Covenant monitoring | Stage 10 | Quarterly scramble for borrower financials | Borrower portal + auto-pulled / parsed statements + covenant calculator |
| LP reporting prep | Stage 11 | Weeks of analyst time per cycle | Templated narrative + auto-pulled portfolio metrics + AI commentary draft |
| Workout data assembly | Stage 13 | Re-reading old emails for context | Searchable deal memory across email, docs, CC minutes |
None of these are speculative; each maps to a vendor or build pattern detailed in 06 · Vendor Landscape and 07 · AI Opportunities.