Entrata AI: ELI+, OXP Studio, And Embedded Agents

How ELI, ELI+, And OXP Studio Fit Together

The names are easy to collapse into one idea because they appear together in product navigation and announcements. That shortcut creates trouble in procurement. A committee may hear “Entrata AI” and assume that every named agent, communication channel, model, dashboard, and automation is part of its existing environment. The sales order and implementation scope often tell a more nuanced story.

Start the evaluation with a product map. Ask the vendor to mark each row as included, separately licensed, dependent on another product, limited by property type, in beta, or unavailable. Then add the internal owner who will configure and maintain it. The map below describes Entrata's public taxonomy as of July 2026; it is a guide to questions, not a substitute for your buying contract.

In some situations, Entrata AI searches may also surface Colleen AI or Redd. Entrata acquired Colleen AI in 2024, and its current public terms retain a separate Redd section. Neither name should be treated as a universal synonym for ELI+, OXP Studio, or the current agent roster. Map any older name to the exact product and workflow in the current proposal.

Map Each Capability To The Work It Performs

The four current ELI+ products cover some of the most frequent resident and prospect workflows. Entrata also describes AI-enabled capabilities in communications, call review, invoice entry, sentiment analysis, lease progression, and accounts payable. Those capabilities may share records or administrative surfaces, but they do not all belong to one package. Keep the product boundary visible while the team evaluates the operating flow. If you're more interested in a specific capability than Entrata's performance in that area, evaluate the leading tools for that function separately.

Use this map as a starting worksheet, then replace every generic phrase with the buyer's own source, output, and owner. “Creates a task” should become “creates a leasing task on the guest card, assigns it to the centralized team within five minutes, and preserves the conversation that caused it.” Precision makes a demo testable and gives implementation a usable acceptance criterion.

Start With High-Volume Intake, Then Test The Exceptions

High-volume intake is a sensible place to use automation. Prospects repeat questions about availability, pets, pricing, tours, and application steps. Residents report common maintenance problems and need to know what happens next. When the source records and policies are current, an embedded capability can collect details consistently, reduce duplicate entry, and place the conversation where the team already works.

And yet, the clean example inevitably tells only half the story. A leasing conversation can shift from an available-unit question to a request for a reasonable accommodation, a complaint about screening, an uncertain fee, or a household situation the approved scripts do not address. A maintenance request can sound routine until a resident mentions smoke, a medical device, a child locked in a room, standing water near electrical equipment, or an inability to follow the suggested troubleshooting steps. Those moments determine whether the workflow supports the staff or merely moves risk faster.

Give Leasing AI One Prospect Who Changes The Script

Run a prospect from first inquiry through appointment and application. Let the initial conversation stay ordinary: the person asks about a floor plan, a move-in date, pet rules, and a tour. Confirm that Leasing AI uses the correct property's availability and policy, writes the interaction to the intended guest card, respects communication preferences, and avoids creating a duplicate when the same person returns through another channel.

Then, change one fact to initiate the curveball. The desired unit is no longer available. The prospect asks whether a fee can be waived, describes an accessibility need, challenges an answer, or begins an application under a different email address. Watch whether the capability recognizes the conflict, preserves the history, stops before making an unauthorized commitment, and routes the issue to a person who can see enough context to continue without asking the prospect to start over.

Do not let a leasing demo end at a booked tour. Ask where the appointment lands, which calendar owns it, what happens after a no-show, how source attribution is preserved, who handles an application that stalls, and what record captures the reason for a handoff. For a wider category view, including the questions that apply beyond one vendor, use our AI leasing assistants for multifamily guide.

Make Maintenance AI Distinguish Urgency From Inconvenience

Begin with a normal request: a dishwasher is not draining. The capability should collect the unit and resident, clarify the symptom, check for duplicate or recent work, offer only approved troubleshooting, capture access instructions, create a useful work order, and set an appropriate priority. The technician should receive something more specific than the original message, and the resident should receive a realistic next step rather than an invented completion time.

Next, use language that can be interpreted more than one way. “The outlet is hot” may refer to temperature, or it could be interpreted to mean the outlet is connected to power. “There is water everywhere” may be a small recurring drip or an active flood. Ask the system to handle a request in another language, then inspect both the resident-facing response and the technician-facing record. Entrata's public terms expressly avoid guaranteeing translation accuracy, so the team needs an emergency rule that does not depend on perfect translation.

The final test should interrupt the normal route. The on-call technician is unavailable, the work order belongs to a vendor, the resident cannot grant access, or the troubleshooting step conflicts with an accessibility or safety concern. Measure how long the request remains unowned, who receives the escalation, whether the resident sees a clear handoff, and whether the source conversation remains attached to the work.

Put Financial And Resident-Retention Workflows Under Tighter Review

Collections and renewals carry operational pressure. Staff need to contact many residents, keep messages consistent, document activity, and focus attention where it can change an outcome. Automation can organize that work, but volume is not the only objective. The communication must reflect the current ledger or offer, respect approved policy, use the right channel, and stop when the resident raises an issue the automated path is not authorized to resolve.

Test Payments AI Against A Ledger Dispute

Use a resident with a straightforward balance first. Confirm the amount, due date, property, account, communication preference, and history that Payments AI receives. Have the capability answer a balance question, issue the approved reminder, and record the exchange. If payment-plan discussion is part of the proposed setup, require the vendor to show where plan rules come from, which terms can be presented, which actions require authorization, and where the result is written.

Then dispute the balance. The resident says a payment cleared, a fee is wrong, a concession is missing, an accommodation applies, or counsel has already been involved. The capability should not keep pressing through a standard sequence. It should identify the boundary, avoid unsupported legal or financial statements, route the issue to the defined owner, and preserve the resident's language alongside the ledger context that triggered the communication.

Inspect timing as closely as wording. A correct reminder sent after a payment posts is still a failure. Ask how quickly ledger changes reach the workflow, what happens during batch processing, how duplicate messages are prevented, and whether a manager can pause communications for a resident, property, or event. The buyer's legal and compliance team should review collection, payment, disclosure, and consent requirements for the jurisdictions in which the workflow will operate.

Make Renewals AI Show The Offer Authority

Renewals AI combines prediction, segmentation, outreach, response handling, and reporting. A committee should look beneath the prediction label. Ask which data points contribute to prioritization, how recently they are refreshed, how a manager can inspect or override the result, and whether the model's output changes the offer or simply the order and manner of outreach.

Run a standard renewal, then add a resident who has a pending transfer, an unresolved service issue, a household change, a disputed notice, a special program requirement, or an offer that needs executive approval. Trace the proposed rent and terms to their source. Confirm who approved them, which versions the resident saw, what happens when the offer changes mid-conversation, and how the system prevents an old offer from remaining active in another channel.

Renewal reporting should also separate contact activity from meaningful progress. A dashboard full of sent messages can conceal stalled offers, unanswered exceptions, inconsistent policy, or residents who received outreach but could not complete the next step. Define success with outcomes the operations and finance teams can reconcile: offers approved, conversations resolved, exceptions aged, renewals executed, notices completed, and records corrected.

Separate The Premium Suite From The Wider AI Feature Set

ELI Call Analysis is one example. The product is presented as a way to transcribe and summarize calls, score them against customer-defined criteria, and support coaching. Its usefulness depends on the quality of the call source, the scorecard and the permissions around sensitive conversations, but its value also depends on the manager's review process. A numerical score should lead back to the transcript and the exact evidence that produced it. The buyer also needs an appeal or correction path when the model misreads a call.

ELI Invoice Entry occupies a different risk profile. Extracting fields from a PDF can reduce repetitive entry, but the work is not complete until the vendor, property, amount, coding, duplicate status, and approval path have been checked. Ask whether the system shows confidence or source location for extracted fields, how exceptions are queued, and whether a draft can reach posting without the required accounting review.

Entrata's June 2026 product materials also describe Resident Pulse, Autonomous Lease Progression, OXP Communications, and AP Payment Agents. Resident Pulse is positioned around sentiment and recurring themes; OXP Communications around a shared workspace for human and agent conversations; Autonomous Lease Progression around orchestration across the leasing lifecycle; and AP Payment Agents around tasks such as voids, reprints, and payment batches. Each deserves its own workflow test. A release-note label does not explain general availability, licensing, eligible property types, source coverage, or the approvals that surround the action.

Autonomous Lease Progression is a useful illustration. Current terms tie it to OXP Studio and active supporting products, with a separate Leasing AI subscription required for autonomous conversation handling. The capability map and sales-order review belong together because the visible experience may feel like one workflow while the commercial and technical boundary spans several products.

Keep a simple register with four columns: capability, product or package, required dependencies, and approved use case. Update it during procurement and implementation. It prevents the committee from promising a future workflow that was shown in a demo but never included in the contracted environment.

Evaluate The Agent Roster One Workflow At A Time

An agent roster is more useful than an agent count. For every proposed agent, record its name, purpose, trigger, source records, required products, action, permission, escalation, write-back, owner, and success measure. Mark whether it is generally available, a limited release, a beta, dependent on another subscription, or restricted by property type. Ask who can activate it and whether activation changes behavior at one property or across the portfolio.

OXP Studio is the proposed control surface for agents, playbooks, escalations, training materials, communications, and workforce activity. A buyer should see those administration screens, not only the polished output of a preconfigured agent. Ask an administrator to change a playbook, revoke access, route an exception, review a conversation, identify a failed action, and export evidence for a monthly operating review.

Use autonomy as a graduated property of a workflow. A capability that summarizes a record and one that initiates a payment-related action need different controls even if both appear in the same roster.

The same agent may operate at different levels in different situations. A leasing workflow might send routine tour reminders but stop at an accommodation request. A payment workflow might answer a balance question but route a dispute. Capture those boundaries as policy, then test them. “Human in the loop” is too vague when no one can identify which human, which queue, or how quickly that person must respond.

Trace The Data From Source To Action

A unified data layer can reduce some handoffs that appear when an external tool must retrieve context through an integration. It still does not prove that a specific workflow has the right context. A property policy may be stored in the platform but outdated. A user may have permission to see a resident record but not to make the action an agent proposes. Two products may display the same concept while refreshing it on different schedules.

Choose a single operating event and keep asking where the next fact comes from. For a leasing inquiry, that may mean availability, rent, fees, pet rules, calendar slots, guest-card history, and application state. For a maintenance request, it may mean the resident, unit, prior work, emergency protocol, access instruction, vendor assignment, and work-order status. The vendor should show the source record, not merely say that the agent “knows” the information.

At every step, ask four questions: Who owns the input? How quickly does it change? Which users and agents can see or alter it? What happens when it is absent or wrong? A workflow should fail visibly and safely. Quietly substituting a generic answer, stale rule, or model assumption can make the experience look smooth while creating follow-up work elsewhere.

Write-back deserves special attention. Determine whether the interaction becomes part of the guest card, resident record, work order, ledger note, task list, communication history, or another system. Confirm that a supervisor can trace the final record to the source conversation and see any human edit. If the team exports data for analytics or compliance review, include that path in the test as well.

Our data readiness for AI in property management guide provides the broader inventory for source systems, documents, permissions, reports, integrations, and data-quality ownership. Use that inventory before configuring an agent to rely on a policy or record that no team maintains.

Read The Product Terms Alongside The Product Demo

Entrata's current public terms repeatedly place operational duties on the customer. Depending on the product, those duties include selecting settings, maintaining complete property information and policies, managing notices and consents, handling escalations promptly, reviewing output, and making sure communications and actions comply with applicable requirements. The terms also acknowledge that AI output can be inaccurate. These are practical design inputs, not boilerplate to leave with procurement.

Translate each responsibility into an owner and a control. “Customer must keep policies current” becomes a named policy library, a quarterly review, an approval step, a publication date, and a test that confirms the agent retrieves the current version. “Customer must handle escalations” becomes a queue, coverage schedule, service target, backup owner, aging report, and resident-facing acknowledgement. “Output may be inaccurate” becomes sample review, exception categories, correction rights, and a threshold that pauses the workflow.

Some workflows touch collections, housing decisions, accommodations, communications consent, consumer information, or other regulated areas. Qualified counsel should determine the legal obligations that apply to the operator's jurisdictions and proposed use. This resource provides an operational review structure; it is not legal advice.

Governance should be sized to the action. A call summary used for coaching and a payment-related communication need different approval, monitoring, and escalation rules. Our AI governance for property management guide explains how to organize policies, permissions, review workflows, pilot controls, incident handling, and an ongoing governance cadence across vendors and use cases.

Ask The Demo To Show The Record, The Exception, And The Handoff

Send the script in advance. A vendor needs enough time to load the right property context and identify features that cannot be shown in the proposed environment. That preparation is useful: it exposes product dependencies before the meeting and discourages a tour of flagship features that never touches the committee's actual work.

Begin after hours with a prospect asking about an available apartment. Have the prospect change the desired move-in date, request a tour, and begin an application. Then leave a required item incomplete. The demo should show which records Leasing AI reads, where the conversation is written, how the appointment is created, how a duplicate is avoided, what Autonomous Lease Progression or another agent does next, and where a person enters when the application remains stalled.

Next, use the same household as a resident. Submit a maintenance request in another language with enough ambiguity to require clarification. Add a phrase that may indicate an emergency and make the normal on-call recipient unavailable. The workflow should translate or structure the request without hiding the original language, apply the emergency policy, create the work order, find the backup route, and preserve a complete handoff.

Move to the ledger. The resident questions a balance after making a payment and mentions a prior concession. Payments AI should retrieve the current context, avoid continuing a standard sequence when the amount is disputed, and route the case to the approved owner. Then open a renewal with an offer exception. Require the team to trace the offer to its source, show the approval boundary, update the conversation when the offer changes, and prevent an outdated version from being sent elsewhere.

Finish in OXP Studio. A manager should find each conversation and agent action, inspect the source and write-back, review an escalation, see which playbook applied, correct a record, change a narrow setting, and explain what would appear in a monthly performance review. Ask the vendor to distinguish live product behavior from roadmap, beta, sample data, and separately licensed capability as it goes.

Score each row as demonstrated, partially demonstrated, described, unavailable, or unresolved. “Described” should never receive the same score as a live result. Keep screenshots or notes tied to the acceptance criterion, then assign an owner and due date to every open dependency. The decision record should survive beyond the people who attended the demo.

Turn The Demonstration Into An Adoption Decision

Select a use case with meaningful volume, a clear owner, accessible source data, and a manageable consequence when the workflow is wrong. Define the properties, channels, hours, roles, and records in scope. List prohibited actions and the conditions that force a handoff. Complete security, privacy, legal, labor, accessibility, communications, and change-management review appropriate to the use case.

• Reconcile the capability map with the signed order, release status, dependencies, and implementation responsibilities.
• Clean and approve the source records, policies, scripts, playbooks, permissions, routing rules, and backup coverage.
• Train users on what the capability does, what it cannot decide, where exceptions appear, and how to correct a record.
• Record a baseline for volume, response time, completion, rework, exception aging, complaints, staff effort, and the business outcome.
• Review a defined sample of ordinary work and every high-risk exception during the pilot.
• Set thresholds for expansion, revision, pause, and termination before results arrive.
• Hold a formal go-forward review with the use-case owner, administrator, operations, and governance participants.

Expansion should follow evidence from the pilot. A capability that performs well at one property may encounter different policies, staffing coverage, languages, property types, or data quality elsewhere. Add scope deliberately, keep the rollback procedure current, and continue reviewing exceptions after the workflow feels routine.

How BC Solutions Supports AI Platform Evaluation

Our work can include current-state workflow discovery, data and permission inventories, buyer-side demo scripts, exception and human-review design, product and dependency mapping, pilot measures, governance roles, training, and change-management planning. We approach the engagement from the operator's side: what work should improve, which records and policies support it, where a person must remain accountable, and how the result will be measured after launch.

For teams already using Entrata, that may mean separating a configuration or process problem from a genuine capability gap before adding another product. For buyers, it means documenting the evidence needed to compare the proposed environment with the current one. Our broader AI adoption enablement services cover strategy, governance, data readiness, workflow design, pilot execution, and organizational adoption across property-management platforms.