How Percent Servicios Inmobiliarios, with 150 employees and a presence across several provinces in Spain and Portugal, unified data, eight channels, 31 automated processes and AI agents in one measurable operating system.
Between February and September 2026, NextAI designed and deployed Sophia for Percent Servicios Inmobiliarios, an organisation with 150 employees, 14 offices and a presence across several provinces in Spain and Portugal. It unified data, acquisition, qualification, assignment, matching, follow-up and management control on a governed architecture.
unified contacts
connected inbound channels
automated processes in production
median time to first contact
new duplicates per week
of repetitive work absorbed
Figures marked “measured” come from system checks and field audits. “Thousands of hours” reflects the updated cumulative scope supplied by the project; it is not converted into financial value or eliminated jobs without specific reconciliation.
Once an organisation reaches a certain scale, the problem stops being a lack of tools. The problem is that no single person can hold the full state of the business in their head. These figures describe the historical perimeter of the ecosystem we worked on.
employees
connected offices
photographs managed
properties in the updated scope
contacts unified in the documented measurement
Updated scope supplied by the project. The initial dossier verified an earlier snapshot of 42,063 properties and 18,534 migrated photographs; the higher figures describe subsequent cumulative growth, not one migration batch.
The business domain, modelled as a connected data structure
Percent Servicios Inmobiliarios has 150 employees and operates across several provinces in Spain and Portugal. Its operation spans connected offices, sales teams and management processes, with years of operational history and a volume of information that grows every day.
The starting point was not a disorganised company. It was a large one. And in large companies the invisible cost is not in doing things: it is in knowing, at any given moment, what state everything is in.
An automation answers a simple, closed question. An enterprise brain answers a question that demands context, judgement and memory. That difference reshaped the project: it was no longer about adding automations around the existing organisation but about revisiting how the organisation should work.
What action should run after this event?
What does this event mean, which state does it change, which process applies, who should step in and what should be remembered for next time?
A linear chain against a network that interprets
The work was not about replacing people but about freeing people from holding the system up. The operation came to rest on five continuous capabilities.
Every interaction, document and business signal becomes structured, comparable information.
States and deadlines stop living in individual diaries: the system creates tasks, detects inactivity and escalates exceptions.
Priorities are calculated on the real state of the business, not on the last conversation somebody remembers.
Processes move forward on their own up to the exact point where human judgement adds value.
Every cycle leaves a trace: what works is consolidated and what fails is corrected in the model, not in someone's memory.
Sophia is not a chatbot added to a CRM. It is a connected operation: 22+ systems, 31 automations and specialist agents work on the same context, with permissions and traceability.
Contacts, properties, prospects and historical operations on one cleaned, shared model.
Eight connected channels with deduplication, assignment, SLA and source traceability.
Carmen answers 24/7, structures conversations and turns 76% of handled calls into leads.
WhatsApp qualifies enquiries within defined limits; the channel was under review at the cut-off date.
An assistant embedded in work screens to query and act on authorised context.
Recalculates enquiry-to-property affinity and presents suggestions in each record.
End-to-end intake, classification, extraction, validation, generation and traceability of business documents.
Reusable automations and bounded services prepared to operate at scale for external customers.
Operational dashboards, a 124-KPI catalogue, seven access profiles and activity records.
Provider-independent architecture: current models can replace earlier generations without redesigning the process.
The AI layer is described by function and capability, not a specific commercial version. This avoids presenting a model name that changes over time as permanent.
An enterprise brain cannot be only a language model, nor only a set of rules. It needs a stable core that guarantees consistency and an intelligent layer that provides interpretation. What is critical always runs the same way; what is ambiguous gets interpreted.
The Percent case combines context and interpretation with explicit rules for the operation.
Diagram based on the description of the Percent case. It does not represent quantitative results or guarantee behaviour outside the scope implemented.
Sophia was documented as a five-layer architecture: data, processes, intelligence, interaction and governance. Value emerges when they work as one system.
One model for contacts, properties, enquiries, opportunities and operations.
Thirty-one production automations maintain states, deadlines, assignment and quality.
Matching and assistants interpret context without replacing critical rules.
Voice, WhatsApp, email, portals and screens operate on the same memory.
Role permissions, evidence, traceability and human escalation support delegation.
Five connected layers, from data to governance
Every action the brain takes is recorded: what ran, with what information, under which rule and with what result. Traceability is what makes it possible to delegate without losing control, and what turns human supervision into an informed decision rather than an act of faith.
Judgement, negotiation and critical decisions remain human. The difference is that people stop spending their time reconstructing the state of things.
The project introduced continuous operational measurement. Improvements are published with their unit and context, without duplicating figures or turning estimates into facts.
The migration did not simply copy files: it rebuilt identities, relationships and history so agents, automations and people could work from one version of the business.
properties · updated scope
photographs managed
contacts · documented measurement
prospects · documented measurement
historical operations
listings in the migration batch
The updated scope (+100,000 properties and +2 million photographs) was supplied after the documentary cut-off. It is distinguished from verified initial-batch figures; personal data, credentials and confidential financial detail are excluded.
Sophia did more than reduce manual tasks. It turned internal capabilities into scalable, reusable infrastructure ready to create services.
Complete systems to receive, classify, extract, validate, connect and generate documents with source traceability and review.
The platform absorbed thousands of accumulated repetitive-work hours and workload equivalent to several roles, without presenting that equivalence as layoffs.
Automations and services created for internal operations were prepared for higher volume and to offer specific capabilities to third parties.
Better acquisition, response speed, portfolio use and multi-channel coverage provided infrastructure to expand reach and market share.
The contribution to market share is described as enabling capacity; no causal percentage is claimed without reconciled commercial measurement.
NextAI combined continuous audit, live verification, system traces, transcribed meetings and owner validation. Each claim is classified as measured data, estimate or context.
normalised evidence records
traced meetings
of transcribed conversation
field audits
Training and documentation were treated as part of the product: role-based journeys, usage criteria and process validation. Economic outcomes remain estimates with explicit assumptions and pending financial reconciliation.
From the map to daily work
This image represents a distributed organisation connected to shared memory. It does not depict real offices or a Percent interface: it visualises the coordination principle developed in this case.
Anatomy of the solution
From input to outcome: explore the four parts of the journey.
Eight channels, calls, messages, portals and documents.
Explanatory diagram based on the described solution. It simplifies the journey; it is not a deployed infrastructure diagram or a live monitor.
Reading the project
In a property network, contacts, more than 100,000 properties, more than 2 million photographs, opportunities and documents belong to the same business. Sophia unified that context, built complete document engines and turned internal automations into scalable capabilities for the network and third-party services.
The architecture combines five layers: data, processes, intelligence, interaction and governance. Deterministic rules maintain consistency; agents handle ambiguous information; traceability preserves human control over relevant decisions.
Connection to today’s offering
The context connecting the company
Explore the serviceCurrent relevance and possible evolution
Percent illustrates an Enterprise Brain in production: Super apps materialise the workspace and Superagents handle specific journeys on shared memory.
This connects the case experience to our current services. Proposed extensions are not presented as features already delivered.
Method and scope
Results compare the same system and publish unit, window and data type. Median time to first contact moved from 68 to 13.2 hours and new duplicates from 64 to 1 per week.
Verified operational metrics and clearly identified estimates are published. Personal data, credentials, sensitive internal rules and confidential financial detail are omitted; results are not a promise for other organisations.
Basis of this analysis: the description, capabilities and stack published in this case. No new measurements are added and no independent audit of its results is implied.
The distinction is not a semantic one. It determines what an organisation can really delegate and what will still depend on somebody remembering.
One isolated task
A complete process with state
None: it reacts to the trigger
The overall state of the business
It fails or stops
It interprets and escalates if needed
It does not remember previous runs
It learns from every cycle
Keeping the automation alive
Setting judgement and supervising
More volume, more maintenance
More volume, more capacity
No company jumps straight to the last level. Maturity is built in layers, and each level is only stable once the one before it really works.
The company uses software. Coordination lives in people.
Specific tasks run on their own, but with no shared context.
The state of the business starts to be single and queryable.
The system proposes, prioritises and prepares; the human decides.
The organisation perceives, coordinates, executes and learns continuously.
Maturity is not bought: it is built in layers
Any competitor can buy the same technology. What they cannot replicate is the accumulated operational knowledge: how this business behaves, which signals anticipate an opportunity, which exceptions recur and how they are resolved. That learning stays inside the brain and compounds over time.
Each cycle feeds the next
Although this case comes from the property sector, the pattern is not sector-specific. Any organisation with volume, distributed processes and decisions that depend on context faces the same problem: the operation grows faster than the human capacity to hold it together.
The next competitive advantage will not be having more tools. It will be building organisations able to perceive, coordinate, decide, execute and learn.
We analyse your real operation — data, processes and decisions — and show you which part could run on its own and which part should stay human.
Eva helps you choose. Our team scopes your project.
We use essential cookies so the site works and, only if you accept them, analytics and campaign measurement cookies. You can change your decision whenever you like. Cookie policy
