LEETConsulting

AI automation · Toronto · since 2018

AI agents that do the work. Built to scale past the pilot.

Leet is a senior team of data scientists, AI engineers and designers. We design, build and run agentic automation for operations-heavy businesses in insurance, healthcare and financial services, where the work has to be right, auditable and secure.

Start a conversation See recent work
We build
Agentic & AI automation
Industries
Insurance, healthcare ops, financial services
Delivery
Pilot to production, then handover
Based in
Toronto, serving North America

Voice agents, document agents, underwriting assistants, workflow automation, evaluation harnesses, RAG and search, forecasting, MLOps.

Why Leet

Most AI pilots stall between the demo and the day-to-day.

The demo works on ten clean examples. Production means messy documents, phone trees, edge cases, audits and a monthly bill.

We have been putting models into real operations since 2018: demand and labour forecasts for national retailers, risk and property data for insurers, analytics for regulated gaming and financial institutions. Agents are the newest thing we build, and we build them the same way: measured against your real cases, observable in production, and designed around your controls.

2018Founded as a data science consultancy. Every year since, models in production.
ProductionEngagements end in running code and a team that knows how to own it.
RegulatedBuilt inside HIPAA, insurance, gaming and financial-services environments.

What we build

From one painful workflow to an automated operation.

We start where the hours go: the calls, the documents, the lookups and the re-keying. Then we automate it end to end, with people in the loop where judgment matters.

01

Agentic workflow automation

Agents that take a case from intake to done: read the file, query your systems, make the call, decide within your rules, write the result back and hand the exceptions to a person.

  • LLM agents
  • Tool use
  • Human in the loop
  • Audit trail
02

Voice agents

Outbound and inbound calls that navigate phone menus, wait on hold, speak with people and log a structured outcome. Every call transcribed and scored.

  • Telephony
  • Speech
  • Post-call QA
03

Document intelligence

Pull facts from PDFs, faxes and forms, check them against policy criteria, and draft letters and summaries for review.

  • Extraction
  • Policy mapping
  • Drafting
04

Underwriting & decision support

Retrieval assistants and research agents that assemble the evidence, so underwriters and analysts decide faster and more consistently.

  • RAG
  • Research agents
  • Risk models
05

Evaluation & AI operations

Test sets built from your real cases, regression runs on every prompt or model change, tracing, and cost and latency you can see.

  • Evals
  • Observability
  • Guardrails
06

Predictive ML foundation

Forecasting, risk scoring, churn and segmentation, delivered as production-ready Python with documentation and knowledge transfer.

  • Forecasting
  • Scoring
  • MLOps

Recent work

Agents in production, under real constraints.

Client names are withheld. The work is described as it was delivered.

2025–26Healthcare operationsAgentic · Voice

Voice agents that sit on hold, so staff don't

For a healthcare services company, we built AI voice agents that call insurance companies, navigate their phone menus by voice and keypad, speak with representatives and record a structured result. Every call is transcribed and scored by an evaluator model, and the outcomes feed post-call analytics.

  • Compliance designed in: covered vendors only, encrypted calling, no sensitive data in identifiers or dev tooling
  • Staged from simulated phone trees to live calls, with a release gate at every level
  • Reports regenerated automatically from transcripts and scores
2025–26Healthcare operationsAgentic · Documents

Document agents for denials and appeals

Agents that read case documents, check them against payer policy criteria and draft appeal letters for staff review. A scored test set decides whether each prompt or model change ships.

2024–25Commercial insuranceLLM · RAG

AI underwriting for an MGA

For a commercial insurance MGA: retrieval-based underwriting support, research agents that assemble web, image and registry evidence into underwriter reports, and a broker workstation from quote to bind.

2019–23Automotive retailForecasting

Forecasts that replaced manual planning inputs

Machine-learning demand and labour forecasts at store level for national automotive-service retailers, so planners start from a model run instead of a spreadsheet. Delivered as production-ready Python with model reports, KPIs and knowledge transfer.

2018–23Insurance dataData pipelines

Automated property and risk data collection

Crawlers, PDF parsing and ML address matching that turned manual research into scheduled pipelines for a national insurance data and analytics provider, feeding its risk models.

2020–22HR technologyNLP

NLP that reads every survey answer

Language models that code free-text survey answers into themes automatically instead of by hand, plus churn models and 360-review analytics feeding a workplace-inclusion recommendation product.

Where we have automated work since 2018

  • Healthcare operationsVoice & document agents
  • Commercial insuranceUnderwriting assistants
  • Insurance dataAutomated data acquisition
  • Agricultural insurancePredictive analytics
  • Automotive retailDemand & labour forecasting
  • Specialty chemicalsPredictive modelling
  • HR technologySurvey NLP & churn models
  • Lottery & gamingForecasting & player models
  • Financial servicesRisk & regulatory analytics
  • EducationData science externships
  • Real estate dataListing & land crawlers
  • Retail & consumerSegmentation & recommendation

How we work

Prove it on your cases. Then scale it.

  1. 01 · Find

    Pick the workflow worth automating

    We map where the hours go, size the return, and agree what "correct" means before anything is built.

  2. 02 · Prove

    Pilot on real cases

    A working agent, measured against a test set from your own cases. Go or no-go is decided on numbers, not on a demo.

  3. 03 · Scale

    Take it to production

    Integrations, monitoring, guardrails and cost controls, running in your cloud and inside your compliance boundary.

  4. 04 · Hand over

    Your team owns it

    Documentation, runbooks and training, so your people can run and extend it. Or we keep running it for you.

Measured, not vibesEvery agent ships with an evaluation set and a score.
People where it mattersClear hand-offs for exceptions and judgment calls.
Your cloud, your dataAWS, Azure or GCP. Nothing leaves your boundary by surprise.
Cost on day onePer-case cost and latency tracked from the first pilot.

Team

One senior team, every discipline an AI project needs.

No pyramid of juniors. The people who scope the work are the people who build it. We assemble the right specialists for each engagement from a bench that has built AI, data and software for insurers, data providers, crown corporations, regulators and global product companies.

AI & ML engineeringLLM agents, retrieval, PyTorch and MLOps on AWS, Azure and GCP. Models shipped into production, not notebooks.
Data science & forecastingForecasting, risk scoring, segmentation and responsible-gaming models, led by former heads of advanced analytics.
Insurance software & integrationTwo decades of broker and carrier platforms: ACORD and CSIO data standards, APIs, identity and security.
Product & UX designDesigners from global agencies and SaaS product teams, so the tools get adopted by the people who use them.
Risk, regulatory & actuarialActuarial, credit-risk and regulatory analytics backgrounds. Controls are designed in, not bolted on.
Strategy & deliveryMBA-trained consultants and delivery leads who keep every build tied to a business case.

AI-native delivery

We use agents to build agents.

Our own delivery runs on coding agents working in parallel lanes, with hard boundaries between regulated and non-regulated data. It is how a small senior team ships at the pace of a much larger one.

It is also the discipline we bring to your environment: clear data boundaries, every action logged, and a human approving what matters.

Our path

From models to agents.

  1. 2018Founded in Toronto

    A data science consultancy for insurance data and analytics.

  2. 2019Retail & industrial ML

    ML forecasting and predictive models for US enterprises in retail, chemicals and agricultural insurance.

  3. 2020NLP & data products

    NLP on survey text, automated property-data pipelines, document parsing.

  4. 2022Production ML & mentoring

    End-to-end ML pipelines; training the next data scientists through externships.

  5. 2024LLMs in underwriting

    Retrieval and research agents for a managing general agency.

  6. 2025–26Agentic automation

    Voice and document agents in regulated healthcare operations.

Insights

Notes from the agentic shift.

All articles
Agents

Your agent's test bench is wired to the real world

Anthropic's test agents filed a false police tip and real web forms. For back-office agents, containment has to start with the first evaluation run.

Read · 5 min Strategy

Don't rent your agents

Gartner predicts 70% of enterprises will abandon agentic AI built by vendor forward-deployed engineers by 2028. With costs falling 13x a year, owning the capability is the whole game.

Read · 2 min Regulation

The AI rulebook timelines moved. Your controls should not

The EU delayed its high-risk AI deadlines, Colorado replaced its AI Act, and insurance regulators are piloting AI evaluation tools. Why we keep building to the stricter standard anyway.

Read · 2 min

Contact

Have a workflow that eats hours every week?

Tell us what it is. We will tell you, honestly, whether an agent can do it, what it would take, and how we would prove it first.

Message us on LinkedIn

Leet Data Consulting Inc. · Toronto, Canada