Kor E
Book a call

Architecture-first consultancy · Austin, TX & Mexico City · Operating across the Americas

We design the system
before we write the code.

Architecture, AI and engineering for banks, retailers, toll operators, hospitals, pharma and logistics companies, among others. Built on the platforms you already run, inside your environment, and owned by you.

Book a discovery callExplore industries 
// ADR-014 · accepted
context  core vendor API, 2 req/s limit
decision async queue + middleware
owner    client · code & data
status   ● approved for build
Our founders have built systems at
  • JPMorgan Chase
  • Walmart
  • MoneyGram
  • Aeroméxico
  • Southwest Airlines
  • Banco Covalto
  • Clara
  • Círculo de Crédito

Industries

Some use cases. One engineering discipline.

Regulated, high-volume businesses where the integration is the hard part. Pick yours to see what we build and how it fits your stack.

Banks · Credit unions · Fintech

Financial institutions

Instant payments, business credit in a day, deposit intelligence and cross-border factoring, built on the core you already run.

Explore Omnichannel · Grocery · Specialty

Retail

One view of inventory, orders and customers across stores, e-commerce and marketplaces, with AI that forecasts and replenishes.

Explore Toll operators · Agencies · Concessionaires

Electronic Toll Collection

Back-office integration, customer service platforms, AI-assisted image review and interoperability for toll operators.

Explore Hospitals · Clinics · Health networks

Hospitals & health systems

Executive and operational intelligence across EHR, ERP and scheduling, with patient flow, revenue cycle and contact-center AI.

Explore Pharma · Clinical research · Distribution

Pharma & life sciences

Clinical operations intelligence, validated data platforms, pharmacovigilance intake and cold-chain visibility.

Explore 3PL · Freight · Cross-border

Logistics

Shipment visibility, cross-border documentation, freight factoring and TMS/WMS integration, with a US–Mexico focus.

Explore

What we do

Three ways we help teams build with confidence.

Primary

Architecture & Advisory

System design, technical due diligence and decision frameworks for complex platforms. We deliver documented architectures that engineering teams can build from, not slides.

  • HLD
  • ADRs
  • LLD
  • Risk matrix

Product & App Engineering

Cloud-native backends, performant web and mobile apps, AI features and integrations that hold up in production.

Team Augmentation

Senior engineers, designers and product leaders embedded in your team. The people in your stand-up are the people doing the work.

How we work

An architecture-first method, applied with consulting rigor.

  1. 01

    Discover

    Goals, constraints and success criteria. We map systems, teams and dependencies to find the real problem behind the symptom.

  2. 02

    Design

    The target architecture, with every meaningful decision documented: HLD, ADRs, LLD and a risk matrix.

  3. 03

    Deliver

    We build, or we embed with your team while you build. Same standards from design to code.

  4. 04

    Evolve

    We support, optimize and adapt the live system through structured maintenance or ongoing advisory.

Built on your core

We connect through your existing platforms and their integration layers. No rip-and-replace.

Inside your environment

Your data stays in your cloud, under your controls and your compliance regime.

Yours to keep

You own the code, the data and the documentation. No auto-renewal, no exit fee.

Selected work

Mobility · Electronic toll collection

A customer service platform for a national mobility operator.

Architecture, documentation and implementation of a customer service platform integrated with the operator's core back office through a custom middleware layer.

9technical documents delivered
12week design phase, kickoff to handoff
✓built and heading into production
See our toll collection work
Customer Service PlatformFreshworks · React SPABUILT
REST · HTTPS
AuthOAuth 2.0
Audit logAppend-only
Custom middleware layerSpring · async queue · RESTDesigned, documented & built by Kor E
Internal API
BOS core systemOperator's back office

Kor E Labs

We build our own products to the same standard.

Two products born inside Kor E, applied to problems we wanted to solve.

VoxStudio

Live · v1.0

AI dubbing platform

Takes source audio, identifies speakers through diarization, transcribes and translates, then regenerates the audio in target voices with AI. Cuts dubbing time and cost by more than 50%.

  • Java
  • Spring
  • Vue
  • PostgreSQL
  • Google Cloud

Korus

In development

AI call center platform

Captures inbound calls, transcribes them in real time and surfaces context to agents, on production-grade telephony and speech infrastructure. Agent assist built for Spanish-speaking markets.

  • Java
  • Spring
  • WebSockets
  • Twilio
  • Google STT

The founders

The team behind the build.

J. Manuel Lomelín

Co-founder & CEO

  • Banco Covalto · fintech architecture
  • Syntage · credit platform
  • Clara · payments infrastructure
  • Círculo de Crédito · regulated services
  • Walmart México · lending platform

Arturo Quiroz

Co-founder & CTO

  • JPMorgan Chase · bank infrastructure
  • MoneyGram · global financial platform
  • Walmart · enterprise architecture
  • Aeroméxico · technology transformation
  • Southwest Airlines · operational systems

Notes

Field notes on architecture and delivery.

AI

AI chatbot or guided options? How to choose

Not every conversation needs a language model. Here's how to decide between an AI chatbot, a bot with buttons and menus, or the hybrid that usually wins.

Arturo Quiroz · Oct 2, 2026 · 7 min read

Read note AI

The future of dubbing

AI is turning dubbing from a premium service for a few titles into a standard step for almost any content. What changes, what doesn't, and what to get right.

Arturo Quiroz · Oct 2, 2026 · 6 min read

Read note Architecture

Why most consultancies skip ADRs (and what that costs you)

ADRs are cheap to write, expensive to skip. Yet most consultancies treat them as overhead. Here's what happens when the design rationale leaves with the consultant.

Arturo Quiroz · May 12, 2026 · 6 min read

Read note Engineering

When to build your own middleware (and when to swallow vendor lock-in)

Custom middleware is the most expensive code you'll ever own, and sometimes the cheapest. Here's how to tell which case you're in before you commit.

Arturo Quiroz · Apr 28, 2026 · 8 min read

Read note

Home / Industries / Financial institutions

Banks · Credit unions · Fintech

Fintech speed.
Earned trust.

Technology and AI for banks, credit unions and fintechs, built on the core you already run and owned by you.

Book a discovery callWhat we build
A historic brick bank branch at dusk with warm light trails passing in front

Who we build for

Three kinds of institution. One engineering team.

Community, regional and commercial banks

Move at fintech speed without replacing your core. We connect through your core provider's integration layer and ship one use case at a time.

  • Send-side FedNow & RTP
  • Small-business credit in a day
  • Deposit attrition early warning
  • M&A data consolidation & Merger Radar

Customers expect real-time finance everywhere. The gap is engineering capacity, and it can be closed.

69%of banks call integration with existing systems a top technology concern
20.7%of small U.S. banks and credit unions are connected to FedNow
1 in 3say their payments offering falls short of customer needs

Sources: Bank Director 2026 Technology Survey · Federal Reserve Bank of Kansas City, July 2026

How we add value

We take what you already do well further.

Bank, credit union or fintech, we start from your strengths, your data and your platforms, and add the engineering capacity to turn them into new capabilities.

Customer relationships

Digital channels and customer data that help your teams know every client better and reach them at the right moment.

Credit decisions

Decision engines that turn cash-flow, tax (SAT/CFDI) and ERP data into same-day answers, under your credit policy.

Speed to market

Senior engineers who ship new products in weeks, on the infrastructure and core you already run.

Data

One real-time data layer across core, cards, lending and acquired systems, so every report starts from the same numbers.

AI

Models and agents that run inside your environment, with your people approving each step.

Compliance & audit

Documented architecture (HLD, ADRs, LLD, risk matrix) that your auditors, examiners and partners can follow.

An analyst walks a banker through a deposit dashboard

Deposit intelligence

Catch deposits before they leave.

Customers keep the account open while their money moves to fintech apps. The warning signs are already in your core: direct-deposit splits, transfers to fintech apps, falling debit-card use and balances eroding over 12 months.

On $1B in deposits, every 10 bps is $1M a year. We deliver a weekly list per branch of who is leaving, how much and where it's going, plus pricing by segment.
Three executives discussing a merger in a boardroom at dusk

Data and integration

A competitor's merger is your opening.

36institutions merged in Q2 2026 alone
11%+average deposit attrition after large-bank conversions

Merger Radar tracks every announced deal in your market, the branches affected and the conversion dates, so your team reaches business clients before conversion weekend.

Sources: FDIC QBP Q2 2026 · Curinos via Bank Director

What we build

Capabilities for financial institutions.

Instant payments

Send-side FedNow and RTP use cases your business clients will actually use: instant payroll, loan disbursement, request for payment. Outbound limits and fraud controls that fit your risk policy.

Business credit in a day

A digital application and a decision engine that uses the cash flow you already see, under your credit policy, with your lenders making the final call.

Deposit intelligence

Direct-deposit splits, transfers to fintech apps, falling debit use. A weekly list per branch of who is leaving, how much and where it's going, plus pricing by segment.

Data and integration

Acquired banks' records moved into one system, or a single data layer if systems stay separate. Merger Radar: every announced deal in your market, so you reach clients before conversion weekend.

Factoring and trade finance

Cross-border factoring for US–Mexico exporters: digital onboarding on both sides, invoices checked against SAT e-invoice (CFDI) records, buyer limits, advances, collections and AML in one flow.

AI that runs locally

Credit models, operations agents and remittance capabilities that run in your own environment. Customer data stays with you.

Reference architecture

How it fits around your core.

Your core & channels

  • Core banking (Jack Henry, Fiserv, FIS)
  • Digital banking
  • Card & ACH processors
  • Loan origination

Kor E build layer

  • Integration & event layer
  • Decision engines under your policy
  • Unified data model
  • AI agents with human approval

What your teams get

  • FedNow & RTP send-side
  • Same-day business credit
  • Weekly attrition list per branch
  • One trusted report

Proof

Fintech builders for banks.

We've built lending engines, credit platforms on Mexican tax data, and payment infrastructure in regulated finance. Now we build them for banks and credit unions.

  • Banco Covalto · fintech architecture
  • Syntage · credit platform
  • Clara · payments infrastructure
  • Círculo de Crédito · regulated services
  • JPMorgan Chase · bank infrastructure
  • MoneyGram · global financial platform

Built on your core

We connect through your existing platforms and their integration layers. No rip-and-replace.

Inside your environment

Your data stays in your cloud, under your controls and your compliance regime.

Yours to keep

You own the code, the data and the documentation. No auto-renewal, no exit fee.

How engagements start

We start with one project and a fixed scope.

  1. 1

    A 30-minute conversation

    Bring the project you've been waiting on. We tell you plainly whether we're the right team for it.

  2. 2

    A fixed-scope pilot

    One use case, one price, running on your data inside your environment.

  3. 3

    Scale it or stop

    Measured against numbers we agree on up front. You keep everything we built.

Other industries

RetailElectronic Toll CollectionHospitals & health systemsPharma & life sciencesLogistics

Home / Industries / Retail

Omnichannel · Grocery · Specialty

Every channel.
One version of the truth.

We connect stores, e-commerce, marketplaces and loyalty into one operational backbone, so inventory is accurate, orders route themselves and AI has clean data to work with.

Book a discovery callWhat we build
A store associate scans a shelf inside a glass-fronted flagship store at dusk, with light trails from passing traffic

The pressure

Where it hurts today.

Inventory you can't promise

When store, warehouse and online stock live in different systems, you oversell online and under-sell in store.

Customers known in pieces

Loyalty, e-commerce and POS each hold part of the customer. Nobody sees the full basket.

Integration debt slows every launch

A new marketplace, payment method or fulfillment option means another point-to-point connection to maintain.

What we build

Capabilities for retail.

Unified inventory & order management

A real-time inventory service and order router that lets you sell what you actually have, wherever it sits: ship-from-store, click-and-collect, endless aisle.

Customer data & loyalty

Customer records matched and cleaned across POS, e-commerce and loyalty into one profile your marketing and store teams can act on.

Demand forecasting & replenishment

Forecasts by SKU, store and week, trained on your sales history, promotions and seasonality, with buyers approving every order.

Embedded payments & credit

BNPL, store credit and instant refunds integrated into checkout, drawing on our lending and payments work in regulated finance.

Integration middleware

One event layer between your POS, ERP, WMS and channels, so the next integration is configuration instead of a project.

Store & back-office agents

AI for price checks, vendor invoices and returns processing, with your staff approving each step.

Reference architecture

How it fits around your core.

Channels

  • POS & store systems
  • E-commerce & app
  • Marketplaces
  • Loyalty & CRM

Kor E build layer

  • Event bus & order router
  • Real-time inventory service
  • Customer 360 data model
  • Forecasting models

Outcomes

  • Ship-from-store & BOPIS
  • Fewer stock-outs
  • Personalized offers
  • Store ops dashboards

Proof

Enterprise retail architecture, from the inside.

Our founders designed enterprise architecture and lending platforms for one of the largest retailers in the Americas. We know what breaks at store scale.

  • Walmart · enterprise architecture
  • Walmart México · lending platform
  • Clara · payments infrastructure

Built on your core

We connect through your existing platforms and their integration layers. No rip-and-replace.

Inside your environment

Your data stays in your cloud, under your controls and your compliance regime.

Yours to keep

You own the code, the data and the documentation. No auto-renewal, no exit fee.

How engagements start

We start with one project and a fixed scope.

  1. 1

    A 30-minute conversation

    Bring the project you've been waiting on. We tell you plainly whether we're the right team for it.

  2. 2

    A fixed-scope pilot

    One use case, one price, running on your data inside your environment.

  3. 3

    Scale it or stop

    Measured against numbers we agree on up front. You keep everything we built.

Other industries

Financial institutionsElectronic Toll CollectionHospitals & health systemsPharma & life sciencesLogistics

Home / Industries / Electronic Toll Collection

Toll operators · Agencies · Concessionaires

Every transaction
accounted for.

We design and build the systems around your back office: customer service platforms, transaction and violation processing, image review and interoperability. Fewer disputes, less leakage, faster service.

Book a discovery callWhat we build
A highway tolling gantry with tag readers and cameras at dusk, with amber light trails from traffic below

The pressure

Where it hurts today.

Revenue leaks between systems

Unread tags, unmatched plates and unreconciled interoperable trips each cost money that rarely shows up in a single report.

Image review doesn't scale

Every plate the OCR can't read goes to a human queue. Volume grows faster than the team.

Customer service sits outside the core

Agents switch between the CRM and the back office to answer one question about a bill, a tag or a notice.

What we build

Capabilities for toll operators.

Customer service platforms

CRM and contact-center platforms integrated with your back-office system through a custom middleware layer, so agents see the account, tag, trips and notices in one place.

AI-assisted image review

OCR confidence scoring that sends only the uncertain plates to people, and learns from their corrections.

Transaction & violation processing

Pipelines that match tag reads and plate images to accounts, apply your business rules and issue notices with a full audit trail.

Interoperability & reconciliation

Clearing and reconciliation with partner agencies, with discrepancies flagged daily instead of discovered at month end.

Payments & account self-service

Autopay, tag top-ups and pay-by-plate flows integrated with your gateway and back office.

AI contact center

Real-time transcription and agent assist for inbound calls, built for Spanish-speaking markets on our Korus platform.

Reference architecture

How it fits around your core.

Roadside & partners

  • RFID tag readers
  • ALPR cameras
  • Interoperable agencies
  • Payment gateways

Kor E build layer

  • Custom middleware (Spring, async queue)
  • Transaction & image pipeline
  • OCR plate review with human QA
  • Reconciliation service

Your back office

  • BOS core system
  • Customer service platform
  • Violations & notices
  • Revenue dashboards

Proof

Designed and documented for a national mobility operator.

We architected and built a customer service platform integrated with the operator's back-office core through a custom middleware layer: nine technical documents, a 12-week design phase and a finished build now heading into production.

  • Freshworks · React SPA front end
  • OAuth 2.0 auth · append-only audit log
  • Spring middleware with async queue
  • HLD · ADRs · LLD · risk matrix

Built on your core

We connect through your existing platforms and their integration layers. No rip-and-replace.

Inside your environment

Your data stays in your cloud, under your controls and your compliance regime.

Yours to keep

You own the code, the data and the documentation. No auto-renewal, no exit fee.

How engagements start

We start with one project and a fixed scope.

  1. 1

    A 30-minute conversation

    Bring the project you've been waiting on. We tell you plainly whether we're the right team for it.

  2. 2

    A fixed-scope pilot

    One use case, one price, running on your data inside your environment.

  3. 3

    Scale it or stop

    Measured against numbers we agree on up front. You keep everything we built.

Other industries

Financial institutionsRetailHospitals & health systemsPharma & life sciencesLogistics

Home / Industries / Hospitals & health systems

Hospitals · Clinics · Health networks

See the whole hospital,
in time to act.

We bring EHR, ERP, scheduling and finance data into one operational picture, so leaders and care teams make decisions on today's numbers rather than last month's.

Book a discovery callWhat we build
Two clinicians and an operations lead review bed-capacity and patient-flow dashboards in a hospital command center

The pressure

Where it hurts today.

Data in every system, answers in none

Census, staffing, supply and revenue each live in their own system. Leadership reports are assembled by hand.

Capacity is managed by phone

Bed availability, discharges and OR schedules are tracked in calls and spreadsheets instead of one live view.

Administrative load on clinical staff

Prior authorizations, documentation and patient calls take time away from care.

What we build

Capabilities for hospitals & health systems.

Executive & operational dashboards

One set of numbers for census, length of stay, staffing, supply and revenue, refreshed daily or in real time.

Patient flow & capacity

A live view of beds, discharges, transfers and OR blocks, with forecasts that flag tomorrow's bottleneck today.

Revenue cycle intelligence

Denial patterns traced back to the documentation and coding steps that cause them, so they're fixed upstream.

Clinical integration

HL7 v2 and FHIR interfaces between the EHR and the systems around it, documented so your team can maintain them.

Administrative AI

Assistants for prior authorization, referrals and inbox triage that draft the work and leave the decision to your staff.

Patient contact center

Scheduling and reminder calls with real-time transcription and agent assist, in English and Spanish.

Reference architecture

How it fits around your core.

Clinical & business systems

  • EHR (HL7 v2, FHIR)
  • ERP & supply chain
  • Scheduling & ADT
  • Billing & claims

Kor E build layer

  • Integration engine
  • Operational data model
  • Alerting & forecasting
  • AI assistants with staff approval

Decisions

  • Executive dashboards
  • Bed & OR capacity view
  • Denial prevention
  • Patient contact center

Proof

Executive dashboards and decision platforms in healthcare.

We've designed executive dashboards and operational intelligence for healthcare and clinical research organizations, built inside the client's environment and under their privacy controls.

  • Executive dashboards
  • Operational intelligence
  • Decision platforms
  • Data stays under your HIPAA controls

Built on your core

We connect through your existing platforms and their integration layers. No rip-and-replace.

Inside your environment

Your data stays in your cloud, under your controls and your compliance regime.

Yours to keep

You own the code, the data and the documentation. No auto-renewal, no exit fee.

How engagements start

We start with one project and a fixed scope.

  1. 1

    A 30-minute conversation

    Bring the project you've been waiting on. We tell you plainly whether we're the right team for it.

  2. 2

    A fixed-scope pilot

    One use case, one price, running on your data inside your environment.

  3. 3

    Scale it or stop

    Measured against numbers we agree on up front. You keep everything we built.

Other industries

Financial institutionsRetailElectronic Toll CollectionPharma & life sciencesLogistics

Home / Industries / Pharma & life sciences

Pharma · Clinical research · Distribution

From trial site
to pharmacy shelf.

We build the data platforms and applications that connect clinical operations, safety, supply and commercial teams, documented to the standard your quality and regulatory teams expect.

Book a discovery callWhat we build
A scientist reviews clinical trial data on a monitor beside sample racks in a laboratory at dusk

The pressure

Where it hurts today.

Trial data scattered across partners

CROs, sites and vendors each report in their own format. Enrollment and site performance are known weeks late.

Every system needs validation

GxP and 21 CFR Part 11 turn every change into documentation work, and many vendors don't produce it.

Supply and safety signals arrive slowly

Temperature excursions, shortages and adverse-event reports move through email and spreadsheets.

What we build

Capabilities for pharma & life sciences.

Clinical operations intelligence

Enrollment, screen-failure and site-performance dashboards that combine EDC, CTMS and vendor data into one study view.

Validation-ready engineering

Our architecture-first method produces design specs, decision records and risk matrices by default, which map directly to validation deliverables.

Pharmacovigilance intake

AI that extracts adverse-event details from emails, calls and forms into structured cases, with safety reviewers confirming every one.

Supply & cold-chain visibility

Shipment, temperature and inventory data from distributors and carriers in one view, with excursions flagged as they happen.

Commercial & field data

Customer and territory data matched across CRM, distributors and market sources, ready for field teams and analytics.

Multilingual content

Training and medical content dubbed and translated with AI on our VoxStudio pipeline, cutting time and cost by more than half.

Reference architecture

How it fits around your core.

Sources

  • EDC & CTMS
  • Safety & PV intake
  • ERP & distribution
  • CRM & field force

Kor E build layer

  • Validated data platform
  • Study & site data model
  • AI case intake with human review
  • Audit trail (21 CFR Part 11)

Outcomes

  • Enrollment & site dashboards
  • Faster case processing
  • Cold-chain visibility
  • Field insights

Proof

Clinical research dashboards and decision platforms.

We've delivered executive dashboards and operational intelligence for clinical research. Every engagement ships with HLD, ADRs, LLD and a risk matrix, the documentation regulated teams need.

  • Clinical research dashboards
  • Documented architecture (HLD · ADRs · LLD)
  • VoxStudio AI dubbing · live v1.0

Built on your core

We connect through your existing platforms and their integration layers. No rip-and-replace.

Inside your environment

Your data stays in your cloud, under your controls and your compliance regime.

Yours to keep

You own the code, the data and the documentation. No auto-renewal, no exit fee.

How engagements start

We start with one project and a fixed scope.

  1. 1

    A 30-minute conversation

    Bring the project you've been waiting on. We tell you plainly whether we're the right team for it.

  2. 2

    A fixed-scope pilot

    One use case, one price, running on your data inside your environment.

  3. 3

    Scale it or stop

    Measured against numbers we agree on up front. You keep everything we built.

Other industries

Financial institutionsRetailElectronic Toll CollectionHospitals & health systemsLogistics

Home / Industries / Logistics

3PL · Freight · Cross-border

Freight moves fast.
Your data should too.

We connect carriers, warehouses and customs into one view of every shipment, automate cross-border paperwork, and get carriers paid faster.

Book a discovery callWhat we build
A dispatcher with a tablet overlooks a logistics yard of trailers at dusk as trucks leave in light trails

The pressure

Where it hurts today.

Visibility stops at the dock

Each carrier reports status differently. Customers call to ask where their freight is because nobody can see it.

Cross-border paperwork in two languages

Carta Porte, CFDI, pedimentos and US customs filings mean documents in two countries, two tax systems and two languages.

Carriers wait to get paid

Small carriers wait 30 to 90 days on invoices and turn to expensive factoring to keep trucks moving.

What we build

Capabilities for logistics.

Shipment visibility

One event hub for carrier APIs, EDI and telematics, normalized into a single shipment timeline your customers can see.

ETA & exception prediction

Models that flag late pickups, dwell time and missed appointments before they happen, so your team calls the customer first.

Cross-border document automation

Carta Porte and CFDI complements generated from shipment data and checked against SAT rules before the truck reaches the border.

Freight factoring & fast pay

Digital carrier onboarding, invoice verification and same-day advances, from the same factoring engine we build for banks.

TMS & WMS integration

Your TMS, WMS, ERP and customer portals connected through one integration layer instead of a web of point-to-point links.

Driver & customer communication

Automated status updates and an AI-assisted contact center for dispatch, in English and Spanish.

Reference architecture

How it fits around your core.

Network

  • Carrier APIs & EDI 214
  • Telematics & ELD
  • WMS & yard
  • Customs brokers

Kor E build layer

  • Shipment event hub
  • ETA & exception models
  • Document automation (Carta Porte, CFDI)
  • Factoring & payments engine

Outcomes

  • Live shipment tracking
  • Proactive exception alerts
  • Faster border crossings
  • Carriers paid in days

Proof

US–Mexico finance and operations, built in.

We've built credit platforms on Mexican tax data and payment infrastructure across the border. Our founders also delivered operational systems for airlines and global payments.

  • Syntage · credit on SAT data
  • MoneyGram · global financial platform
  • Aeroméxico · technology transformation
  • Southwest Airlines · operational systems

Built on your core

We connect through your existing platforms and their integration layers. No rip-and-replace.

Inside your environment

Your data stays in your cloud, under your controls and your compliance regime.

Yours to keep

You own the code, the data and the documentation. No auto-renewal, no exit fee.

How engagements start

We start with one project and a fixed scope.

  1. 1

    A 30-minute conversation

    Bring the project you've been waiting on. We tell you plainly whether we're the right team for it.

  2. 2

    A fixed-scope pilot

    One use case, one price, running on your data inside your environment.

  3. 3

    Scale it or stop

    Measured against numbers we agree on up front. You keep everything we built.

Other industries

Financial institutionsRetailElectronic Toll CollectionHospitals & health systemsPharma & life sciences

Home / Notes

AI

AI chatbot or guided options? How to choose

Not every conversation needs a language model. Here's how to decide between an AI chatbot, a bot with buttons and menus, or the hybrid that usually wins.

by Arturo Quiroz · Oct 2, 2026 · 7 min read

Almost every client conversation about customer service now includes the same request: “we want an AI chatbot.” Sometimes that's exactly right. Often, what the business actually needs is a bot that offers clear options, follows a fixed path and never improvises. Choosing wrong is expensive either way: a guided bot frustrates customers with open questions, and an AI bot in a regulated transaction creates risk you didn't need.

Two different tools

A guided bot (options, buttons, menus, decision trees) offers the customer a set of choices and moves through a flow you designed in advance. Every answer it gives was written and approved by someone on your team.

An AI chatbot uses a language model to understand free text and generate a response, ideally grounded in your own documents and data. It handles questions you didn't anticipate, in the words the customer actually uses.

When a guided bot is the right call

  • The intents are few and well known. Check a balance, pay a toll, reschedule an appointment, track a shipment. If 80% of contacts fall into a dozen intents, a menu resolves them faster than a conversation.
  • The wording is regulated. Disclosures, consent, KYC steps and collections scripts must say exactly what compliance approved. A guided flow guarantees that.
  • The steps have a strict order. Payments, transfers, claims and returns need validation at every step. Buttons and forms make that sequence explicit and auditable.
  • You need predictable cost and behavior. A guided bot costs the same per conversation and behaves the same every time. It's easier to test, certify and explain to an auditor.
  • The channel favors taps over typing. On WhatsApp and in mobile apps, quick-reply buttons are often faster for the customer than writing a sentence.

When an AI chatbot earns its place

  • Questions are open and varied. “Can I use my card in Colombia if I'm on a payment plan?” has too many variations for a tree. A model grounded in your policies can answer it.
  • The knowledge base is large. Product manuals, benefit plans, tariffs and policies that no menu could cover without becoming a maze.
  • Customers mix languages and phrasing. Spanish, English and everything in between. Language models handle this far better than keyword matching.
  • Context matters. Summarizing a customer's history, explaining a bill line by line or preparing an agent before they pick up the call.

The hybrid usually wins

In most of our projects the answer is both. The AI understands what the customer wants and answers informational questions. The moment the conversation turns into a transaction, it hands off to a guided flow with fixed steps, validations and approved wording. Money never moves on the strength of a generated sentence.

Three guardrails make the hybrid work:

  1. Grounding. The model answers from your documents and data, with sources, and says “I don't know” when the answer isn't there.
  2. Escalation. A clear path to a human agent, with the conversation summary handed over so the customer doesn't repeat themselves.
  3. Logging. Every prompt, response and handoff recorded, so compliance can review what was said and why.

Five questions before you choose

  1. What are the top 20 reasons customers contact you, and what share of volume do they cover?
  2. Which of those involve money, personal data or regulated wording?
  3. How often do your policies and products change?
  4. Which languages and channels do your customers use?
  5. Who reviews conversations, and how will you measure resolution?

If the top intents cover most of your volume and involve transactions, start with a guided bot and add AI for the long tail. If the questions are open and the knowledge is deep, start with AI and route transactions to guided flows. Either way, design the handoffs first. That's where most bots fail.

This is the same thinking behind Korus, our AI call center platform: real-time transcription and agent assist where language is open, and fixed, auditable flows where the business requires them.

Book a discovery callAll notes

Notes

The future of dubbingWhy most consultancies skip ADRs (and what that costs you)When to build your own middleware (and when to swallow vendor lock-in)

Home / Notes

AI

The future of dubbing

AI is turning dubbing from a premium service for a few titles into a standard step for almost any content. What changes, what doesn't, and what to get right.

by Arturo Quiroz · Oct 2, 2026 · 6 min read

For most of its history, dubbing has been reserved for content that could justify the cost: feature films, hit series, major campaigns. Everything else (training videos, product demos, internal communications, back catalogs, podcasts) stayed in its original language or got subtitles at best. AI is changing that math.

What the pipeline looks like today

A modern AI dubbing pipeline has five stages, and each one has improved sharply in the last few years:

  1. Diarization. Identify who speaks when, so each voice is handled separately.
  2. Transcription. Turn speech into text with timing, including names and terminology.
  3. Translation and adaptation. Translate meaning, not words, and fit the line to the time available on screen.
  4. Voice generation. Produce the new audio in a voice that matches the speaker's tone, pace and emotion.
  5. Mix. Put the new voices back over the original music and effects so it sounds like it was recorded that way.

We built VoxStudio around exactly this pipeline. In production it cuts dubbing time and cost by more than half.

Where it's going

  • The long tail gets dubbed. When cost drops, the question changes from “is this title worth dubbing?” to “why isn't everything available in the viewer's language?” Corporate training, e-learning, medical education and catalog content are the first to benefit.
  • Voices keep their identity. A speaker can be heard in Spanish, English or Portuguese with their own voice, with consent. That matters for executives, instructors and creators whose voice is part of the message.
  • Emotion and timing improve. The hard part is no longer generating a clean voice but carrying over emphasis, pauses and feeling. That's where most current progress is.
  • Video follows audio. Lip-sync adjustment for on-camera speakers is moving from research into production tools.
  • Near real time. Live events, webinars and calls with simultaneous translated audio are becoming practical, with a delay of seconds rather than days.

What doesn't change

People still direct the work. A dubbing director or reviewer decides how a joke lands in another culture, whether a term should be translated at all and when a line needs rewriting. AI does the heavy lifting; people make the calls that define quality.

Consent and rights come first. A voice belongs to the person who owns it. Any serious pipeline needs explicit consent for voice cloning, clear terms for how a voice can be reused, and records that prove both. Actors, studios and platforms are already setting these rules, and they should shape how the technology is built.

Localization is more than language. Units, currencies, references and regulations change across markets. Medical and financial content in particular needs subject-matter review before it ships.

What to get right if you're adopting it

  1. Start with content where speed and reach matter more than artistic nuance: training, product and internal communications.
  2. Build a glossary of names and terms before the first run. It's the cheapest quality improvement available.
  3. Keep a human review step and measure how many edits it requires. That number tells you when the pipeline is ready to scale.
  4. Get voice consent in writing and store it with the project.
  5. Keep your source files and outputs in your own environment, especially for confidential or regulated content.

Dubbing is becoming infrastructure: a step in the publishing workflow rather than a separate production. The organizations that treat it that way will reach audiences in their own language by default.

Book a discovery callAll notes

Notes

AI chatbot or guided options? How to chooseWhy most consultancies skip ADRs (and what that costs you)When to build your own middleware (and when to swallow vendor lock-in)

Home / Notes

Architecture

Why most consultancies skip ADRs (and what that costs you)

ADRs are cheap to write, expensive to skip. Yet most consultancies treat them as overhead. Here's what happens when the design rationale leaves with the consultant.

by Arturo Quiroz · May 12, 2026 · 6 min read

Most architecture handoffs we see go like this: a deck with diagrams, a Notion page with bullet points, and a meeting where someone explains the “why” out loud. Three months later, the team is rebuilding half of it because nobody remembers why a key decision was made.

This is what Architecture Decision Records (ADRs) exist to prevent.

What an ADR actually is

An ADR is a short document, usually one page, that captures a single architectural decision. The format varies, but the useful version always includes four parts:

  • Context. What problem are we solving? What constraints are real?
  • Decision. What did we decide?
  • Consequences. What does this commit us to? What did we give up?
  • Status. Proposed, accepted, deprecated, superseded by ADR-XXX.

That's it. The whole document is rarely longer than a page. The discipline isn't in the format. It's in writing it down before everyone forgets the trade-offs.

Why consultancies skip them

Three reasons, in roughly this order:

They take time to write. A real ADR, one that captures the actual reasoning and not a sanitized version, takes 30-60 minutes per decision. On a 12-week engagement with 8-12 meaningful decisions, that's a day of work. Most consultancies don't include that day in their estimates.

They make the consultant replaceable. This is the uncomfortable one. A well-documented architecture is one a different team can pick up and extend. If your competitive advantage is being the only person who knows why things work the way they do, ADRs threaten that.

They show your reasoning. Some decisions look obvious in hindsight but were uncertain at the time. ADRs preserve that uncertainty: the alternatives considered, the trade-offs accepted. That transparency makes some consultancies uncomfortable.

What it costs the client

A client who doesn't get ADRs from their architecture engagement is buying a system they can run but not evolve. Three to six months in, when the team faces its first real fork (a new integration, a scale problem, a regulatory change), they're flying blind on every decision the original architect made.

What ends up happening:

  • Someone “discovers” why a decision was made by breaking something.
  • The team starts over on parts of the architecture because they can't verify whether the original logic still holds.
  • A second consultancy is hired to re-document what the first one delivered.

Each of these is expensive. None of them appear on the line item of the original engagement.

How we use ADRs at Kor E

Every architecture engagement we run ships with an ADR log. The format is the four-part one above. We write them as we go, usually in the same session where the decision is made, while the reasoning is still fresh.

We commit them to the client's repo, not ours. They belong to the client. We've had engagements where the client's next engineer arrived six months after we left and built a feature correctly on the first attempt because they read the ADR for the decision they were extending.

That's the whole point.

Book a discovery callAll notes

Notes

AI chatbot or guided options? How to chooseThe future of dubbingWhen to build your own middleware (and when to swallow vendor lock-in)

Home / Notes

Engineering

When to build your own middleware (and when to swallow vendor lock-in)

Custom middleware is the most expensive code you'll ever own, and sometimes the cheapest. Here's how to tell which case you're in before you commit.

by Arturo Quiroz · Apr 28, 2026 · 8 min read

Every architecture engagement involving system integration eventually arrives at the same question: do we build a custom middleware layer, or do we accept the vendor's integration surface as it is?

The wrong answer is expensive in both directions. Let's break down when each is right.

What “middleware” means here

By middleware I mean a service or layer your team owns that sits between two systems and translates between them. It might handle authentication, retries, format conversion, queueing, throttling, or business logic that doesn't belong in either system.

The alternative is to let the systems talk to each other directly using whatever the vendor provides (SDKs, webhooks, REST endpoints) and absorb the limitations of that interface into your application code.

When to build it

There are four signals that justify custom middleware:

You need to change behavior the vendor doesn't expose. If the vendor's webhook fires once and your business requires three retries with exponential backoff plus an alert at retry two, and the vendor doesn't support that, you need a layer that does.

The protocol mismatch is structural. If your platform speaks events and the vendor speaks request/response, every consumer of that integration is going to pay the cost of bridging that gap. Centralizing it in one place is cheaper than 12 partial implementations.

You expect to switch vendors. A custom middleware abstracts the vendor away. If your CRM, billing, or telephony provider is likely to change in the next 24 months, the middleware is what saves the migration.

You need audit, compliance, or observability beyond what the vendor offers. Regulated industries often need every call to be logged, signed, or routed through a specific region. Middleware is where that lives.

When to skip it

Custom middleware is one of the most expensive pieces of code you'll ever own. It needs to be deployed, monitored, scaled, secured, and updated whenever either side changes. If none of the four signals above apply, skip it.

Specifically, skip it when:

  • The vendor's interface already meets your needs end-to-end.
  • The integration is one of many low-volume ones where the cost of owning a middleware exceeds the cost of duplicating client logic.
  • The vendor is genuinely strategic to your business and switching is implausible (you're integrated with Stripe and you're not leaving Stripe).
  • Your team doesn't have the bandwidth to operate the middleware as production infrastructure.

A real example

On a recent engagement, the client integrated their customer service platform (Freshworks) with their core operational system (a custom BOS). Freshworks had no native concept of the BOS's primary entities, and the BOS had no concept of tickets.

A direct integration would have meant every Freshworks plugin or workflow needing BOS data would have to know the BOS's API, auth model, and rate limits. We modeled it and counted twelve points of integration coming in the next year.

The decision was easy: a custom middleware layer that exposed a stable, opinionated API to Freshworks and translated calls to the BOS underneath. The middleware also handled retries, audit logging, and a queue for async operations the BOS couldn't process synchronously.

Cost: one engineer for 12 weeks. The payoff: every future integration point is designed to ship in days, not weeks, and a future vendor change becomes a contained problem.

The decision framework

Before you commit, write down:

  1. What capability does the middleware add that you can't get from the vendor?
  2. How many consumers will use the middleware?
  3. What happens if you don't build it? What's the duplicated cost?
  4. Who operates it after you ship?

If you can't answer #4, you shouldn't build it. A middleware without a team to operate it is technical debt with a uniform.

Book a discovery callAll notes

Notes

AI chatbot or guided options? How to chooseThe future of dubbingWhy most consultancies skip ADRs (and what that costs you)

Let's work together

Let's find your first project.

Thirty minutes. Bring the project you've been waiting on, whether it's on this site or not. If it runs on your platforms and your data, we can scope it. We reply within one business day.

J. Manuel LomelínCo-founder & CEO, Kor E Consulting jmlr@kor-e-consulting.tech

Offices

Austin, TXUnited States+1 (512) 222-3074
Mexico City (CDMX)Mexico+52 55 3219 4092