JF Lefebvre Consulting

Power BI and workflow automation

Stop rebuilding Monday's report.

I'm J-F. I build Power BI dashboards and workflow automations, then hand them over so your team can keep them running.

Let's work together

Contract or permanent. Based in Edmonton, open to on-site, hybrid and remote work in Canada.

IMonday, by hand

  1. Open last week's workbook. Save it as v8.
  2. Pull three exports. One site is spelled four different ways.
  3. Paste them together. The pivot breaks.
  4. Hunt for the duplicate rows by eye.
  5. The totals don't match Finance. Find out why.
  6. Rebuild the chart. Again.
  7. Send it, late, with an apology. See you next Monday.

Here's the same report once it's set up.

IIThe workbook

Same three exports. The cleanup is written once and runs every week after.

  1. Make site names and dates agree
  2. Drop the duplicate rows
  3. Group and total by site
  4. Model it with payroll hours
  5. Publish it: refreshed every Monday, 06:00

Sample data: a made-up incident log and payroll hours for three work sites. The cleanup, model, rates and dashboard are computed live in this page.

Sample incident log, summarised by site

IIIThe dashboard

An executive summary that opens onto the accounts behind it.

This is a sample mortgage portfolio, generated in this page: 12,000 accounts at month end and twelve months of history. Filter it by insured status or arrears measure, then pick a cell on the heat map to open the accounts inside it.

In Power BI the second view would be a drill-through page. This one is plain HTML, so you can try it without a licence.

Month end · Mar 2026

Arrears by province and LTV

Share of balance in arrears. Pick a cell to see its accounts.

Arrears 90+, 12 months

Sample data only: made-up accounts generated by the script on this page, with arrears rates set well above a real Canadian book so the heat map has something to show. No client data.

Built to be handed over

The model
A current-month account table and a monthly history table, joined through shared Date, Geography and LTV band dimensions.
The measures
Balance, weighted LTV and arrears rates, each written once and documented, so a new visual doesn't need a new formula.
The refresh
History grows a month at a time. The account snapshot is replaced at each month end, so the model stays the same size.
The walkthrough
A session on how the pages, measures and refresh fit together, and a written guide for whoever maintains it next.
// Share of balance in arrears, for any slice
Arrears 90+ % =
DIVIDE (
    CALCULATE (
        [Balance],
        fact_Accounts[Days Past Due] >= 90
    ),
    [Balance]
)

Weighted LTV =
DIVIDE (
    SUMX (
        fact_Accounts,
        fact_Accounts[Balance] * fact_Accounts[LTV]
    ),
    [Balance]
)

IVWhat I do

Power BI reporting, and automating the work around it.

Power BI development

Executive summary pages, drill-through to the rows behind a number, slicers, heat maps and scorecards. Underneath, a star schema and DAX measures sized for the data, so the report stays quick and the next person can follow how it works.

Power BI · DAX · Power Query · SQL · Excel

Workflow and AI automation

Report runs automated with Power Automate. A Python pipeline that pulls from several sources, scores the results with an LLM and sends a daily brief. AI screening in a live recruiting platform, where a person reviews every decision. Each one needs a log and an exception path, because sooner or later it meets data it wasn't built for.

Power Automate · Python · SQL · LLM APIs · PostgreSQL · Docker

People and safety analytics

Headcount, turnover, overtime and incidents usually live in separate systems. Put them side by side and you can see where a short crew or a long stretch of overtime starts costing lost days. I spent six years in workforce analytics after starting out in transportation compliance, so I know what these exports look like when they land.

HRIS extracts · Turnover and overtime · Incident scorecards

VTrack record

Six years of workforce and people analytics, at Clean Harbors and Graham. Since 2024 I've also built and run software of my own.

  • 2019

    Incident trends tracked by hand, region by region.

    Clean Harbors. Excel and Power BI scorecards for incident trends by region, and PeopleSoft extracts turned into expiring-documentation reports that operations managers acted on.

  • 2022

    Weekly, monthly and quarterly packages pulled together from HR, Finance and IT systems.

    Graham Management Services. Self-serve Power BI dashboards presented to executives, built with SQL and DAX across several sources. Report runs automated with Power Automate, and workforce models that improved headcount planning accuracy.

  • 2024

    A 350+ member trading community depending on one set of tools.

    MQL4 Expert Advisors for MetaTrader 4, built under a service agreement and kept running with version control, a bug-fix pipeline and structured support.

  • 2026

    Recruiters screening résumés after hours.

    BuildFlow Workforce, an AI recruiting platform in production since June 2026: AI screening under mandatory human review, scheduling and reporting, on PostgreSQL and Docker. Alongside it, jobfinder, a Python pipeline that scores job postings with an LLM and sends a daily email brief.

VIHow I work

I start with the people who use the report or run the process. What decision does it feed, which numbers do they already trust, and what does the data look like when it actually arrives? I started as a material handler and worked up to shift lead, and watching the work before changing it is a habit from then.

Then we agree on the few views or steps that matter most, and cut the rest to fit the time and the budget. I'd rather ship a smaller report that gets opened every week. Data gaps, access limits and other risks go on the table in the first week, not the last.

Handover is part of the job. Your team gets a walkthrough of the model, the measures and how it refreshes, plus a written guide, so it doesn't need me to keep running.