Multiglass

One business, a dozen tools, and no single source of truth.

I clean up the foundation and give your team one place to run the work — starting wherever it hurts most.

The shape of the work

Most teams run their day across eight tabs, three Slack threads, and a spreadsheet that nobody owns. Multiglass starts wherever it hurts most — cleaning up the data, automating the rote work, retiring the tools that have become a tax — and, once the foundation underneath is interoperable, builds the operator portal that ties it together. Discovery, design, and engineering by the same hands.


Services

  • Data cleanup & migration

    Spreadsheets that disagree. Exports that take a day. Reports nobody trusts. I get the data into a state where decisions can rest on it.

    What an engagement looks like

    A two- to three-week engagement landing on one reliable source for the numbers your team disagrees about. Usually starts by shadowing whoever pulls the weekly report by hand.

    • A canonical schema for the entities the team argues about
    • An idempotent ingest pipeline from your current upstream sources
    • A small regression suite that flags schema drift before reports break
    • A one-page reference for non-technical staff
  • Workflow automation

    Re-keying the same fact in three places, copy-pasting between tools, weekly reports built by hand. The rote work comes off the team.

    What an engagement looks like

    Identify the three or four manual processes eating the most operator time. Automate them in order of pain. Often delivered as a small set of scripts plus a thin internal interface for kicking them off.

    • A short discovery doc enumerating every place data is re-keyed
    • Targeted automations for the worst offenders, with a logged audit trail
    • A thin internal UI for one-off runs and parameter tweaks
    • Runbooks for the edge cases that still want a human in the loop
  • Stack modernization

    Tools that have turned into a tax instead of an advantage. Replace what is past its useful life with something smaller, built for the work the team does now.

    What an engagement looks like

    Pick the one tool that has become the biggest tax — a CRM nobody trusts, a spreadsheet that should be a database, a homegrown app no one wants to touch — and replace it with something scoped to today’s needs, not yesterday’s.

    • A migration plan with a no-data-loss guarantee
    • The replacement, deliberately smaller in surface area than what it replaces
    • A parallel-run period before cut-over
    • The old system retired cleanly, with an exit-data archive the team can read
  • Systems integration

    Sales sees one number, finance sees another, and nobody is wrong. Make the systems you keep speak the same language.

    What an engagement looks like

    Define the canonical model of the entities that span systems — customer, order, invoice, asset. Build an integration layer that keeps them in sync: event-driven where it matters, scheduled where it does not.

    • Typed contracts for each cross-system entity
    • A webhook + queue layer for real-time syncs
    • Reconciliation jobs that catch the inevitable drift
    • A small dashboard showing the health of every integration
  • Unified operator portal

    A single pane of glass — one surface for the people running the business. Earned once the foundation underneath is interoperable.

    What an engagement looks like

    Built only after the data and integration foundation underneath is solid. The portal is the easy part once everything underneath agrees with itself. Usually six to ten weeks for v1, scoped to a specific operator role.

    • A single console for one operator role, expanded later as it earns the right
    • Live data from the integrated foundation — no fresh silo created
    • The five to ten most common actions a few clicks away
    • Role-based views and a changelog the team trusts
  • Decision support & ML

    Forecasting, classification, anomaly detection, routing. When the data is clean and connected, models earn their keep.

    What an engagement looks like

    Applied where a clear decision the business already makes can be improved. Always paired with the integration and data work that makes it sustainable — a model on a swamp is just an opinionated swamp.

    • A baseline model with accuracy claims you can check and a confusion matrix you can read
    • A feature pipeline built on the integrated foundation, not a one-off spreadsheet
    • A retrain pathway that does not require me to be in the room
    • A small service that exposes predictions to the rest of the stack

Process

  1. 01

    Listen

    A short engagement shadowing the people who do the work.

  2. 02

    Map

    Workflow diagrams, friction points, a scope with no padding.

  3. 03

    Build

    Designed and engineered by the same hands.

  4. 04

    Hand off

    You own it. Retainer is optional.


Who I help

Concrete by role. Find the one that sounds like your week — or tell me which I am missing.

Non-profits

  • Executive Director

    Stretched between operations, fundraising, and program oversight — with no time to step back.

    A single dashboard for org health (program, finance, development, HR), built on cleaned-up data. Less time assembling reports, more time leading.

  • Program Director / Manager

    Tracking participants, outcomes, and grant deliverables across spreadsheets and a CRM that nobody loves?

    A program-management surface that produces the reports your funders expect, automatically. One place for participants, outcomes, attendance, and case notes.

  • Development / Grants

    Re-keying donor data, building grant reports by hand, chasing outcome numbers from program staff every quarter?

    Integrate the CRM, accounting, and program systems so grant reports come from one source of truth. The narrative is yours; the numbers assemble themselves.

  • Operations / Admin

    HR, IT, finance, facilities — and the tools that should make it easier are themselves the work.

    Cut the rote work. Replace the tools that have become tax. A short, honest engagement that takes the worst handful of pain points off your plate first.

  • Finance / Bookkeeping

    Reconciling between bank, donation platform, accounting, and program budgets every month?

    An integration layer that makes monthly close honest — donations, restricted vs. unrestricted, program spend, payroll, all reconciled to one ledger.

  • Communications / Storytelling

    Writing donor updates and grant narratives whose data is always slightly out of date?

    Build the pipes so your story has the latest numbers automatically — outcomes, beneficiaries served, dollars raised, all queryable when you draft.

  • Volunteer Coordinator

    Sign-ups in one tool, scheduling in another, hours in a third, recognition in a spreadsheet.

    Consolidate into one volunteer surface — sign-up to thank-you note, with hours and roles reportable at the org level.

  • Board / Governance

    Quarterly board packets assembled by hand from disparate systems, outdated by the time they ship?

    A board dashboard the ED can produce on demand. Same numbers, every quarter, with drill-down for the questions that always come up.

Small & medium businesses

  • Founder / Owner

    You know exactly what is broken. Every fix means one more tool to manage.

    Pick the worst three operational pains. Build a small, focused interface for the team that retires more tools than it adds.

  • Operations Lead

    The canonical "truth" of any given fact depends on who you ask.

    Stand up the integration and the operator portal that put one number in front of the whole team. Built around your actual workflow, not the tools.

  • Sales / Account Manager

    CRM that tells you nothing useful, deal-stage updates that happen in three different threads?

    A pipeline view that pulls from the systems where work actually happens. Less data entry, more selling.

  • Customer Success / Support

    Switching between ticketing, the product database, and the billing system to answer a single customer question?

    A single customer surface — history, billing, current state — built for the people who answer the phone.

  • Finance / Bookkeeping

    Manual journal entries every month from systems that should already be reconciling themselves?

    Build the integrations between billing, payments, payroll, and your accounting system. Close shrinks from a week to a day.

  • IT / Technical Lead

    You are the only one who knows how it is all wired. No time to write it down.

    A second pair of senior hands — discovery, integration, modernization, and the documentation that means it does not all break when you take a vacation.

Mid-market companies

  • COO / Director of Operations

    Operators running the day across too many tools, with the cost showing up as context-switching and slow onboarding?

    A unified operator portal — built on top of the integration and data work that makes it possible — scoped to a specific operator role for v1, widened as it proves out.

  • IT / Engineering Director

    A backlog of integration and modernization work that never reaches the top of any sprint?

    Senior contract help that hands off cleanly. Discovery, design, implementation, and the documentation your team can actually maintain.

  • Department Heads (Sales, Service, Finance)

    Reports that take a week to produce, numbers that disagree with the next department, decisions made on hunches?

    Integrate the systems you already pay for. Build a department-level dashboard that reflects the same source of truth as everyone else.

  • Data / Analytics Lead

    Modeling on top of pipelines that break weekly, while stakeholders ask for ML before the data can support it.

    Clean the foundation, build the integration layer that makes the data trustworthy, then put the models that pay for themselves on top.


About

Multiglass is run by Thomas Kavanagh, a Portland-based engineer with a decade of background in data science and applied ML. Discovery, design, and engineering by the same hands — mine, end to end.

Start with a discovery call.

Thirty minutes. Tell me where the work piles up, and I will tell you where I would start.