For listed REITs and their managers, commercial landlords and residential developers. Built to the SEBI REIT Regulations, NDCF, Schedule V, LODR and BRSR, or to the listing regime that applies to you.
Every figure in every report links back to the page and the cell it came from.
In brief
EigenSpark builds AI systems, and the engineering around them, for listed REITs, REIT managers and real estate developers. The document work is lease abstraction and gap analysis, insurance comparison at renewal, valuation reports read into the Schedule V sheet, and the annual report reconciled against the signed financials. The platform work is the lease and asset master, the golden source under it, and integration with the systems you already run. One REIT engagement started with a single finance use case and became a shared platform carrying three.
Regulatory frameworks
These decide what a listed REIT publishes and when. Below is our work under each one, in the order a REIT meets them.
Amended 23 March 2026
We build against the rule set you report on, and we make it editable. The March 2026 amendments changed what surplus can move up from SPV and HoldCo level and what has to be disclosed separately once NDCF is computed. Checks like that change; a rebuild should not.
With results, half-yearly and annual
We automate the assembly. The judgement stays with your team. Rent, opex, interest, capex and distribution figures are pulled from the source workbooks, matched by meaning, so a renamed line still lands, and every figure that does not reconcile across sheets is flagged before the pack goes out.
Mandatory minimum disclosures
We read the valuation report and fill the Schedule V sheet, recording the source page for every field. A second check maps your statements against the applicable SEBI, LODR and ICAI requirements and reports what is missing or incomplete, linked back to the rule.
Top 1,000 listed entities, filed annually
We build the data path. Your team writes the narrative. Meter readings, tenant allocation, water and waste records and green certifications are pulled from the building management system and the utilities layer into the BRSR format, carrying the same source trail as a financial figure. Assured reporting and green bond covenants both need that trail.
We build the checks and the evidence trail. We do not sign your disclosures. That judgement stays with your compliance officer, your valuer and your auditors. Outside India we map the same controls to your own listing regime.
The starting position
Shifting labels, a close calendar that does not move, figures spread across SPVs, and drafts read by hand.
The workbook, the signed financials, the notes and the annual report each describe one line item their own way. A system that matches on exact text fails on the second document. Matching has to work on meaning, and it has to survive a change of nomenclature.
Results, the NDCF statement, the half-yearly and the annual report land on fixed dates. Since 1 January 2026 mutual funds treat listed REITs as equity, and index inclusion opened on 1 July 2026. More institutional holders asking harder questions, on the same dates.
REIT, HoldCo, SPV, asset, building, unit. A number is only right at the level it belongs to, and the eliminations between levels are where the errors hide. Any check worth running has to know which level it is looking at.
A revised lock-in, an altered escalation clause or a reduced sub-limit gets through on the fourth read of a negotiated draft. Nobody finds out until the clause is invoked, which can be years later.
How an engagement runs
Real estate engagements start with a build. Scope on your documents, ship one use case, then reuse the layer underneath it.
We work from your real leases, valuation reports and signed financials. Two to three weeks with your finance, legal or leasing team to agree the fields, the rules and what a correct answer looks like on your documents.
A single use case taken all the way to a working web interface: upload, extract, review, download, audit trail. Around forty person-days for a first module, sized to what the documents actually demand.
Ingestion, extraction, semantic matching and the rule engine are shared, so the second and third use cases cost a fraction of the first. New rules and checklist items get added by your team without a release.
Use cases
Leasing, finance and disclosure, property operations, then investor relations and developer sales.
Every executed lease read into a structured record: term, escalation, lock-in, renewal and termination options, deposits, CAM and revenue share. WALE, expiry ladders and option calendars then compute off one source.
Document Intelligence Engine →A negotiated draft compared against your standard template. Every change to rent, escalation, security deposit, CAM, termination, subletting and indemnity, tagged by severity and shown beside the clause it replaced.
Contract Risk Analyser →Lease milestones, payment behaviour, service-ticket themes and space usage read together into a health score per tenant, with expansion and exit signals surfaced early enough for the asset team to act on them.
Agentic AI →A hundred-page valuation PDF read, the mandatory minimum disclosure fields extracted and normalised, and the sheet returned filled, with the source page recorded per field and a review step before anything is accepted.
Generative AI →Standalone and consolidated figures matched by meaning, across the annual report, the signed financials and the audit report. Every mismatch reported with its page, table and figure number, and re-run until it clears.
Continuous Compliance Monitor →Crawlers watch SEBI, MCA and ICAI for new circulars. A person confirms what applies, the master list updates, and your statements are checked against it. Gaps come back linked to the regulatory source.
Agentic AI →Lease schedules, escalations, renewals, expected vacancy and debt outflows read into a rolling thirteen-week cash view. Rate resets, delayed receipts and tenant default are tested against it, with DSCR and refinancing triggers flagged before treasury is cornered.
AI & Data Strategy →Budget against actual across rental income, operating cost and capex, with a plain-language explanation written for each movement and every figure linked to its source. The draft management commentary arrives with the variance.
Generative AI →Utility meters, the building management system, tenant allocation and waste and water records assembled into the BRSR format on a schedule, with the source trail an assurance provider or a green bond covenant asks for.
Data Engineering & Platforms →Chillers, lifts and power systems watched for the pattern that precedes a failure. Recurring fault tickets grouped, work orders drafted with the vendor history attached, contractor response times tracked, and HVAC and lighting tuned against actual occupancy.
Predictive Maintenance System →An assistant that answers routine investor and analyst questions from published disclosures and board-approved material, cites the source, and routes anything sensitive to investor relations before it goes out. Quarterly factsheets assemble from the same numbers.
Generative AI →A name and a city enriched from public sources into a profile: company roles, directorships, wealth signals, confidence on the match. Sales gets a ranked list with the evidence behind each rank.
Lead Enrichment Agent →The engineering half
A model is the short part of a REIT project. These four layers decide whether any of it survives your close calendar.
One record for tenants, leases, assets and the chart of accounts, versioned, with a full audit trail underneath.
Most of the effort in a REIT project sits here.
We have written the functional architecture for a REIT ERP and we build against it, one module at a time, around what already works.
Defined once and used everywhere, so the factsheet, the board pack and the investor answer cannot disagree. Your team runs and owns it.
We also run the training that goes with it: SQL and data engineering, cloud, enterprise systems and cybersecurity, alongside the AI tracks. A team that cannot query its own data cannot specify a model either.
What we offer real estate and REIT teams
Most engagements here start with one document use case and end up drawing on three of the four.
Where most REIT engagements start.
The AI build, and the engineering under it.
AI tracks, and the foundations under them.
Working in a specific function? See how we help Finance & Accounting teams.
Security and controls
A human review step, the source recorded for every figure, rules your own team can change, and a full audit trail.
Every extraction and every flagged deviation goes to a person before it counts. The system proposes and evidences; your finance, legal or leasing team decides.
A number in a report links back to the page, table and cell it came from. An auditor can follow the trail without asking us or waiting for us.
Verification rules, compliance checklists and the regulatory sources being watched are managed in a console. When SEBI amends something you do not wait for a release.
Every run, decision, rule change and user action is recorded, with each document versioned as it was uploaded. The record matches what an audit asks to see.
FAQs
In your tenancy or on your hardware, whichever you run. A typical deployment is an application and API instance, a persistent volume for the working database, and object storage for the uploaded documents and the generated output, all inside your own cloud account. Where a hosted model is used for extraction, we agree in advance what is sent to it and what never leaves. On-premise with open-weight models is a standard choice here and it holds up in production.
Accurate enough to be worth reviewing, which is the standard that matters. We set the target on your own reports during scoping. Long documents are handled in stages, with confidence scoring per field, tables and scanned pages routed differently from body text, and every field carrying the page it came from. Anything below threshold is surfaced for a person to decide.
It breaks systems that match on exact text, which is why ours do not. Line items are matched semantically, so depreciation described four ways across the workbook, the notes, the fixed asset register and the annual report still reconciles as one item. Label detection is dynamic, so a restructured sheet does not need a code change.
The work is sequenced so the checks exist before the pack does. Consistency checks run on the workbook as it is built, the annual report comparison runs as drafts circulate and re-runs until it clears, and disclosure gap checks run against a master list your team keeps current. The point is to find a mismatch in the week it is created.
Around forty person-days for a first module, after two to three weeks of scoping on your documents. That covers ingestion and parsing, the extraction pipeline, field mapping and validation, output generation, the review interface, deployment, testing, user acceptance and handover. The second use case is materially cheaper because the platform under it already exists.
No, and we would argue against it. The systems of record stay. We build the modules that are missing, integrate what already exists, and put a golden source underneath so the lease master, the chart of accounts and the asset hierarchy agree. Where a REIT ERP module genuinely does not exist for your structure, we build that module.
No, and the parts that are not are usually what makes the AI work. Integration, the data model, the golden source, access control and workflow are ordinary engineering, and they are most of the effort. The same applies to training: alongside the AI tracks we run SQL, data engineering, cloud, enterprise systems and cybersecurity, because a team that cannot query its own data cannot specify or judge a model either.
The document work applies wherever contracts, policies and vendor bids are read by hand. The sales work is different: enrichment turns a thin lead record into a profile your team can rank, with the evidence attached, which matters most where site visits are the constraint. Bid and tender evaluation for fit-outs and AMCs sits between the two.
Related
The two products this work usually ships as, the service behind them, and one REIT case study.
The REIT finance team, the disclosure pack, and what shipped.
→ ProductLease and policy comparison, clause by clause.
→ ProductDisclosure checks against a rule set your team can edit.
→ ServiceThe accounting system, the PMS, the registrar and the BMS.
→ FunctionThe function that owns most of this work in a REIT.
→ PillarThe engineering side, for the systems that go into production.
→Send us one negotiated lease, one policy renewal, one valuation report or one disclosure pack. We call you within 48 hours and go through what can be extracted reliably, what has to stay with a person, and what a first module would take.
Talk to us