Custom web appsAI toolsStrategy

WhitepaperCustom software vs. SaaS

Custom software for your toughest problems.

LekkerLogic is a small, hands-on team that builds custom web apps, AI tools and strategy around the way your business actually works. No one-size-fits-all software.

Animated examples of LekkerLogic automation: reviewing a sales inbox to log each email in Salesforce, reading a PDF purchase order into SQL, cross-referencing a competitor part number with AI, and analyzing pricing history to recommend price changes.

Built for teams that make, sell, distribute and service real products

01Selected work

Custom-built. In production. Doing real work.

Each of these started as a client's toughest problem. None of them came from a template, and every screen shows sample or public data.

01

Project delivery platform

TaakBoard

A project management platform for teams that sell, specify and deliver custom-manufactured work. Committed ship dates, free manufacturer access, an Outlook add-in and an AI briefing of what's at risk this week.

  • Project management
  • AI Insights
  • Outlook add-in
taakboard.com (opens in a new tab)
02

B2B customer portal

Hoffman Portal

Where nVent Hoffman customers look up parts, see their own pricing and live factory stock, pull drawings and CAD, and turn a parts list into a quote. Browsing patterns feed recommendations and inventory forecasting.

  • Customer portal
  • AI assistant
  • Analytics
hoffmanportal.com (opens in a new tab)
03

Sales intelligence platform

Visie iQ & Visie.ai

A custom CRM and sales intelligence platform with interactive dashboards, AI account reviews, predictive models, Outlook add-ins and a secure partner portal.

  • Custom CRM
  • Sales intelligence
  • Forecasting
visie.ai (opens in a new tab)
04

CRM integration API

Visie iQ API

A secure REST API that keeps Visie iQ opportunities in step with each manufacturer's own Salesforce or Microsoft Dynamics 365. Changes flow both ways, each partner sees only its own lines, and CRM updates never loop back.

  • REST API
  • Salesforce
  • Dynamics 365
05

Team work-style assessment

TeamDNA

A work-style assessment used by teams where people answer short questions in about ten minutes and get a report on the kinds of work that energize and drain them. Leaders see the whole group as one team: who works well together, where the gaps are, and how a new line-up would change the mix.

  • Work-style assessment
  • Team insights
  • Laravel 12
06 Confidential client

AI meeting assistant

Meeting Notes

07

Live flight tracking

Flight Tracker

Also built

  • Email-to-order entry

    Customers email a PDF purchase order. AI extracts every line, matches the customer and manufacturer, and queues it for a quick check before it's saved.

  • Event lead capture

    A no-code form builder with printable QR codes. At trade shows, visitors fill it in on their phone or snap a business card, and AI reads it into the CRM.

  • Quote and parts-list builder

    Paste rough quote text or a screenshot of a parts list. Get exact part numbers and order quantities from your catalog, ready to export.

  • Competitor cross-reference

    Turn a competitor's part number, a customer's parts list or a spec sheet into your matching products, with the differences flagged for review.

  • AI briefings and business reviews

    One click writes an executive briefing or a customer business review from live data, with follow-up chat and a branded presentation export.

  • Secure partner portals

    Vendors, dealers or customers get their own sign-in to see the projects and data that involve them, and nothing outside their scope.

02Why custom

Off-the-shelf software was built for someone else's business.

Generic platforms get you most of the way. Then your team fills the gap with spreadsheets, workarounds and copy-paste. That gap is where your toughest problems live, and it's exactly where we work.

Read the whitepaper: custom software vs. SaaS
  1. Off-the-shelf:Bent to fit the toolThe real work still happens in a spreadsheet beside it.
    Built for you:Built around your processNot the other way around.
  2. Off-the-shelf:Re-typing what's already knownEmails into the CRM, orders from PDFs, reports by hand.
    Built for you:Busywork, automatedCustom tools and AI agents do it. Your people approve.
  3. Off-the-shelf:AI without a planNobody knows where it pays off, or who would build it.
    Built for you:AI where it pays offWe find it, plan it and build it with you.
  4. Off-the-shelf:Someone else's roadmapYou pay per seat and live with it.
    Built for you:Yours, and connectedMade by the people you meet, wired into your CRM, ERP and inbox.

03What we do

Custom solutions, shaped around your business.

Every engagement starts with your problem, not our product. Then we build what it needs: a web app, an AI tool, an integration, a plan, or all of them together.

  • 03.1

    Strategy & planning

    We learn your pain points, rank the opportunities, including where AI really helps, and map out a plan you can act on.

  • 03.2

    Custom web apps

    Portals, CRMs, dashboards and internal tools designed around how your team actually works, not how a vendor thinks it should.

  • 03.3

    AI tools & agents

    Emails become CRM notes, PDFs become orders, and pricing gets reviewed overnight. Anything consequential waits for a human click.

  • 03.4

    Systems integration

    CRM, ERP, inbox and spreadsheets, finally talking. Enter data once and see it everywhere.

  • 03.5

    Data & reporting

    Dashboards that answer plain-English questions, plus forecasts for demand, margin and churn.

  • 03.6

    Training & support

    Hands-on training for the people who'll use it, clear guidelines for AI, and a team that sticks around after launch.

04How we work

Consult, Plan, Build, Deliver.

One small team, with you from the first conversation to launch day and beyond. You always know where things stand, and you always talk to the people doing the work.

  1. Stage 01

    Consult

    We sit with your team to understand the pain points, the workflow and the data behind it, and agree on what "solved" looks like.

    A shared definition of solved
  2. Stage 02

    Plan

    A clear plan with scope, priorities, timeline and the result we'll be measured on.

    A plan you can act on
  3. Stage 03

    Build

    Short cycles with regular check-ins, so you see progress and steer as we go. Integrated with the systems you already run.

    Progress you can steer
  4. Stage 04

    Deliver

    Launched, documented and your team trained. We stay on to support it and improve it as your business changes.

    Launched, trained, supported

05How we build AI

AI you can trust with real business data.

AI should take busywork off your team's desks, not replace them. Your proprietary data is what makes it useful, so how that data is handled is part of the design, not an afterthought.

Every AI feature follows the same path
  1. Messy inputEmail, PDF, photo
  2. AI interpretsStructure from chaos
  3. Code verifiesRules and math
  4. Person approvesOne quick check
  5. System updatedCRM, ERP, SQL

Data governance

Your data stays your data.

Your data reaches AI only through business APIs that don't train on it, kept for the shortest time possible and limited to what each task needs, under the same access rules as your people. We document how it all works in plain language for your IT and legal teams.

  • P1

    The right model for each job

    Top models for deep analysis, fast low-cost ones for bulk reading. Quality where it counts, predictable costs everywhere else.

  • P2

    AI interprets, code decides

    AI turns messy text, photos and PDFs into structure. Tested code does the matching and the math.

  • P3

    People decide, with sources in view

    Nothing touches a customer, a price or a record until someone approves it, and every answer shows its source.

  • P4

    Instant answers, not repeat bills

    Results are cached until new data arrives, so you never pay twice for the same analysis.

  • P5

    Secure by design

    Scoped access from day one, locked-down public pages and rate-limited AI on public forms.

  • P6

    No lock-in to one AI vendor

    Models are set in configuration, so moving to a better or cheaper one is a settings change, not a rebuild.

Tools we build with

  • Claude API
  • OpenAI API
  • AI agents & tool use
  • Python
  • Node.js
  • TypeScript
  • PHP
  • MySQL & SQL Server
  • REST APIs
  • OAuth 2.0
  • Salesforce
  • Microsoft Dynamics
  • Outlook add-ins
  • Webhooks
  • Live dashboards

06Start a conversation

Bring us your toughest problem.

It starts with a conversation about your pain points. We'll tell you honestly what we'd build, how we'd build it and what it would take.

  1. 01A 30-minute call with the people who'd build it
  2. 02Your pain points, mapped and prioritized
  3. 03A clear plan for what we'd build and how
new-project.requestReply by email

Please enter your name.

Please enter a valid email address.

Please tell us a little about the problem.

No spam. We only use your details to reply.

Case 01 / 07TaakBoard
TaakBoard

Client buildProject delivery platform

TaakBoard

A project management platform for teams that sell, specify and deliver custom-manufactured work. It tracks every project, manufacturer and ship date from purchase order to the delivery truck.

Type
Project management SaaS, built for a client
Built for
Fabricators, OEMs, integrators, reps and dealers
What we did
Product design, build, AI and integrations
Replaced
A Monday.com setup that never fit the workflow
taakboard.com
TaakBoard dashboard with the active pipeline, overdue projects and a ranked list of projects that need attention

01The challenge

The team ran projects on Monday.com, but a general-purpose board never fit how custom-manufactured work moves. Manufacturers, POs, lead times and committed ship dates had to be forced in with workarounds, so making it sort of work took more effort than it should have for what it cost. Ship dates still drifted, decisions got buried in email, and "where are we?" took a round of calls to every manufacturer.

02What we built

A custom platform built around their workflow and one question: will it ship on time? Manufacturers, POs, lead times and ship dates are built in rather than bolted on, so the team works the way it already does instead of bending to a generic tool.

03The outcome

One dashboard ranks what needs attention. Committed dates stay on the record when they slip, and manufacturers check status themselves instead of waiting on a phone call.

Highlights

  • Committed ship datesWhen a date moves, the promised date stays beside it and the slip is flagged. Nothing is quietly overwritten.
  • Free manufacturer accessSecure read-only links let manufacturers see their projects with no account and no paid seat. Links can expire and every visit is logged.
  • Outlook add-inSave an email, its attachments or a task to the right project without leaving Outlook. AI can summarize a long thread first.
  • AI InsightsA prioritized briefing of what's urgent, what's at risk and which manufacturers need a nudge this week.
  • CRM and data connectionsWon orders arrive from Salesforce, Dynamics and other CRMs, and a Data API feeds ship dates to the ERP.
  • Every view a team needsKanban, calendar, timeline and map views, plus documents, tasks and @mentions on every project.

From order to delivery

  1. Your CRMOpportunity won
  2. TaakBoardProject created, ship date projected
  3. OutlookApproval email saved to the project
  4. TaakBoardShipped on the committed date
Case 02 / 07Hoffman Portal
nVent Hoffman

Client buildB2B customer portal

Hoffman Portal

Where nVent Hoffman customers go from part number to quote in one place: live factory stock, their own company pricing, drawings and CAD, and quote requests for a whole parts list.

Client
JD Martin Company
Built for
nVent Hoffman customers: distributors, integrators, OEMs, contractors and engineers
What we did
Portal design, build, AI assistant and CRM sync
hoffmanportal.com
Hoffman Portal home page with a part lookup showing factory stock, company pricing and CAD downloads

01The challenge

Customers had to call or email for stock, pricing, drawings and quotes. Every answer meant waiting on a person, and every question landed on the sales team.

02What we built

A portal tied to each customer's company: search any part, see factory stock and their negotiated price, pull every document, and send a full parts list for a quote.

03The outcome

Customers help themselves around the clock, and JD Martin's sales team receives complete, priced parts lists instead of piecemeal questions. Browsing patterns feed recommendations and forecasting.

AI in the product

How the AI finds the right part

Customers rarely know the exact part number. AI turns what they do know into the same requirements the configurator asks for, then the portal's own matching code picks real, active catalog parts. The AI never invents a part.

  • Claude Sonnet 5.5
  • Structured JSON outputs
  • Rule-based matcher
  • Keyword fallback
  1. Enclosure Assistant
    Plain language to requirements

    inStainless wall-mount, about 24 × 20 × 8, for a washdown area

    outNEMA 4X · stainless · wall · 24 × 20 × 8 in

    The assistant reads a description and fills in the configurator for the customer: NEMA rating, material, mounting and size. It marks each requirement as stated or assumed from the application, and follow-up messages refine the same search.

  2. Suggested enclosures
    Requirements to ranked parts

    inNEMA 4X · stainless · wall · 24 × 20 × 8 in

    outClosest-fit enclosures at the customer’s own price

    Stated requirements filter and assumed ones only rank. NEMA ratings must meet or exceed the request, sizes are ranked by closeness and widened when nothing fits exactly, and every result shows the company’s own pricing.

  3. Competitor part numbers
    Unknown number to Hoffman equivalent

    inA competitor number missing from the cross-reference table

    outBrand, size, rating and material, plus Hoffman candidates

    The AI decodes the number using that brand’s known cross-references as examples. Only active catalog parts are kept, results are labelled as unconfirmed AI suggestions, and every miss is logged so the table grows from real lookups.

Highlights

  • Real-time inventoryStock at the factory warehouses on any part, with an estimated ship date.
  • Company-specific pricingEach customer sees its own negotiated price on every part and project, not just list price.
  • Every part documentedSpec sheets, 2D and 3D CAD, manuals and compliance documents on one page per part.
  • Projects and quote requestsBuild a parts list per job with the team, send it for a quote and follow it from Requested to Quoted.
  • Three ways to find a partA requirements configurator, a competitor cross-reference and an AI assistant that takes a plain-language description.
  • Secure company accountsEvery account is tied to its company, so users see only their own pricing, projects and quotes.

From part to quote

  1. SearchFind the part by number, spec or AI
  2. ProjectAdd it to a project at your pricing
  3. Quote requestSend the whole list in one request
  4. QuoteReview the quote in the portal
Case 03 / 07Visie iQ & Visie.ai
Visie

Client buildSales intelligence platform

Visie iQ & Visie.ai

A custom CRM and sales intelligence platform with interactive dashboards, AI account reviews, predictive models, Outlook add-ins and a secure partner portal.

Client
JD Martin Company
Type
Custom CRM and sales intelligence, built for a client
Built for
JD Martin's sales team and partners
visie.ai
Visie.ai home page: "Turn sales data into growth opportunities" beside an account dashboard and a morning briefing of flagged declines

What it does

  • Interactive dashboards
  • AI account reviews
  • Predictive models
  • Outlook add-ins
  • Secure partner portal
  • Custom CRM
Case 04 / 07Visie iQ API
Visie

Client buildCRM integration API

Visie iQ API

A secure REST API that connects Visie iQ to each manufacturer partner's CRM, whether that's Salesforce or Microsoft Dynamics 365. Opportunities flow out as they change, wins and updates flow back, and every partner sees only its own lines.

Client
JD Martin Company
Connects
Visie iQ with each manufacturer's Salesforce or Microsoft Dynamics 365
What we did
API design and build, access control, sync design, and integration guides for both CRMs
Stack
PHP, Slim 4, MySQL, JSON over HTTPS
Visie iQ API
Diagram: the Visie iQ API between Salesforce and Dynamics 365. Changed opportunities are pulled out to each CRM, and status, amount, dates and notes are written back. Every request is checked against the partner's own key, permissions, manufacturer scope and daily limit, and logged, and CRM updates never echo back.

01The challenge

JD Martin's team works opportunities in Visie iQ across many manufacturers, and each manufacturer runs its own CRM. Some use Salesforce, some use Dynamics 365. Both sides needed the same pipeline without re-keying it, and each manufacturer could only be allowed to see the deals that involve its own products.

02What we built

One REST API on Visie iQ, plus an integration guide for each CRM. Every manufacturer gets its own key, limited to its own manufacturer lines and to read-only or read-write access. CRMs pull only what changed and send wins, amounts, close dates and notes back.

03The outcome

Manufacturers can work JD Martin's opportunities for their lines inside the CRM they already use, and updates made there show up in Visie iQ on the same record. Changes that come from a CRM never echo back to it, and every call is logged and visible on a usage dashboard.

Highlights

  • A key for every partnerEach manufacturer gets its own key, limited to its own lines and to read-only or read-write access. Anything outside that scope is refused and logged, and each key has a fair-use limit.
  • Only what changedEach CRM picks up just the opportunities that changed since it last checked, so syncs stay small and fast and nothing re-sends the whole pipeline.
  • No sync loopsUpdates that come from a CRM are never sent back to it. Only changes people make in Visie iQ go out.
  • Safe write-backPartners can update only a few approved fields on their own opportunities, and every change is saved to the opportunity's history. New opportunities arrive complete, with their notes, owners, companies and contacts.
  • The same record on both sidesEach opportunity is linked across both systems, so an update in either one lands on the right record instead of creating a duplicate.
  • Integration guides for both CRMsA guide for each manufacturer's Salesforce or Dynamics 365 team takes them from setup to their first sync, so they can connect without a support call.

From a Visie iQ update to the CRM and back

  1. Visie iQOpportunity updated and flagged to sync
  2. Delta pullOnly changed records for that key
  3. Salesforce or DynamicsLands on the matching record
  4. Write-backWin, amount and dates return, with no echo
Case 05 / 07TeamDNA

Client buildTeam work-style assessment

TeamDNA

A ten-minute work-style assessment that turns individual answers into a clear picture of the team. Each person gets a report on the work that energizes and drains them, and leaders see the whole group on one page.

Built for
Leaders, managers and HR teams who want better conversations about how people work
What we did
Product design, scoring engine, reports and team views
Framework
Based on concepts from Patrick Lencioni's The 6 Types of Working Genius. Independent, and not affiliated with The Table Group.
TeamDNA
TeamDNA report for a sample person: their pairing, The Idea Champion, a short summary and their six types marked energizing, workable or draining

01The challenge

Teams rarely talk about the kind of work that drains people, so the wrong person ends up owning the launch, the follow-through or the review, and nobody can say why things keep stalling. Off-the-shelf personality tests hand back a label, not who should pair with whom on this team.

02What we built

A questionnaire anyone can take from a group link on their phone, with no account. Scores combine 36 ratings with 15 head-to-head choices, and close calls go to a short tiebreaker the person ranks themselves. The report appears the moment they finish, arrives by email, and joins the group’s team view.

03The outcome

Each person gets a plain-language report: their pairing, their two energizing, two workable and two draining types, and questions to bring to the conversation. Leaders see the whole team on one page, compare any two people, and try a new line-up before making the change.

Highlights

  • Ten minutes, on any phoneFifty-three short questions in two parts from a group link. No account and no cookie: a private link is the only key, and answers save as you go so people can pause and come back.
  • Scoring that holds upRatings and head-to-head choices across all 15 pairs of types are combined. Reverse-worded and check questions catch rushed answers, and close scores go to a tiebreaker the person ranks.
  • A report people actually readTheir pairing, six types in three bands, where they add the most value, watch-outs and questions to talk through. Print or save it as a PDF, with a private link by email.
  • The team on one pageA strand map of who is energized and drained by each type, a people-by-type grid, and gap and balance warnings for the group.
  • Pairs and line-upsCompare any two people side by side, see who works well together, and build a line-up to see how it changes the mix before you reorganize.
  • Roles and privacySuper Admins, Admins, and Group Admins who see only their own groups. Consent is recorded with every result, and there is no tracking or analytics.

From group link to team view

  1. Group linkOne link shared with the team
  2. Questionnaire53 questions, about ten minutes
  3. ReportShown on finish and sent by email
  4. Team viewEveryone in the group on one page
Case 06 / 07Meeting Notes
Confidential client

Client buildAI meeting assistant

Meeting Notes

A private alternative to AI note-taking bots. Nothing joins the Teams or Zoom call: after the meeting, upload a recording or transcript and get the summary, decisions and action items with owners, then search or ask questions of any meeting later.

Client
Confidential client
Built for
Leaders and sales teams who spend their day in meetings
What we did
Product design, build, AI pipeline and search
AI models
OpenAI API: GPT-6.1 Terra for notes and Q&A, OpenAI embeddings for search
Private client workspace
Meeting note with a summary, key points, the recording and details of who attended

01The challenge

Decisions lived in whoever took notes, and follow-ups slipped. Popular AI note takers weren't an option: their bots join Teams and Zoom calls automatically from the calendar, announce themselves to clients and record conversations no one meant to share. In client meetings, that cost more trust than the notes were worth.

02What we built

A sign-in-only app that stays out of the meeting. The team records the way they already do, or keeps the transcript Teams or Zoom already makes, and uploads it afterward. Through the OpenAI API, the app transcribes it, has GPT-6.1 Terra write structured notes, and indexes every meeting with OpenAI embeddings so the whole library can be searched by meaning.

03The outcome

Clients never see a bot on the call, yet every meeting still ends with notes in the same format. Every action item has an owner and a date, and anyone can ask "what did we decide?" and get an answer drawn from the transcript.

Highlights

  • No bot in the meetingNothing auto-joins a Teams or Zoom call, nothing announces itself to your clients, and nothing listens unless you choose to upload the recording afterward.
  • Recordings or transcriptsMP3 or M4A files up to 1 GB from a phone, Zoom or Teams, or a transcript you already have. Long meetings are split up and transcribed piece by piece.
  • Structured notesGPT-6.1 Terra writes a summary, decisions, open questions and action items with an owner and a due date. Email them to the team in one step or copy them anywhere.
  • Ask about any meetingAsk what someone said or why a decision was made, or have it draft a follow-up. GPT-6.1 Terra answers from the transcript, and "Expand on this" digs into any point.
  • Search across every meetingOpenAI embeddings index the notes and transcript of every meeting, so a search like "meetings about pricing" finds matches by topic even when the words differ. Exact names and phrases match too.
  • Private by designSign-in only. Meetings live in the team's own workspace, not a note-taking vendor's, and recordings are encrypted at rest and play through links that expire.

From meeting to searchable notes, with no bot on the call

  1. MeetRun the call as usual. No bot joins
  2. UploadRecording or transcript, afterward
  3. GPT-6.1 TerraNotes, decisions and action items with owners
  4. EmbeddingsIndexed for search across every meeting
Case 07 / 07Flight Tracker

Client buildLive flight tracking

Flight Tracker

Live flight tracking built on the OpenSky Network API, with Denver International (KDEN) as the hub. The team sees every departure, arrival and aircraft on the ground at Denver, follows any flight by its callsign, and looks up where a flight number has been flying.

Hub
Denver International Airport (KDEN)
What we did
Product design, build, API integration and flight logic
Data
OpenSky Network REST API: live ADS-B positions and flight records
Private client app
Denver hub board: a map of aircraft across the United States beside a live list of departures from KDEN with climb rates, ranges and flight levels

01The challenge

The team needed to see what was moving at Denver International right now and follow specific flights, without paying for a commercial flight-data feed. The OpenSky Network has the data, but its live feed carries no origin or destination, its flight records arrive a day late, it can't be searched by callsign, and every query spends from a daily credit budget.

02What we built

Three connected views on one server-side API layer: a Denver hub board that sorts traffic into departures, arrivals and aircraft on the ground, a flight tracker that follows up to six flights by callsign, and a live map of the Front Range. The server holds the API credentials, caches every answer and shares it with every open tab.

03The outcome

Departures are confirmed by seeing an aircraft on the Denver surface and then in the air, and arrivals by a range to the field that keeps closing. Likely destinations come from where the flight number went on recent days and are labelled as inferred. A hundred open tabs cost the same API credits as one.

Highlights

  • Denver hub boardEvery aircraft within 120 nm of KDEN, sorted into departures, arrivals and on the ground. Altitude rules are measured above Denver's 5,434 ft field, not sea level, so short final never reads as cruise.
  • Live flight trackingFollow up to six flights by callsign or flight number. Each gets its own color, a trail, an altitude profile, speed, climb rate and departure airport, with FAA VFR and IFR charts as an optional overlay.
  • Callsign lookupsOpenSky can't search by callsign, so the tracker finds the aircraft once with a worldwide sweep, remembers its address, then polls a small box around its last position for a quarter of the credits.
  • Routes and destinationsFlight records show where each flight number went on previous days. Answers that point at mountain airstrips after a lost signal are filtered out, and what remains is shown as inferred, never as fact.
  • Secure API accessOAuth2 client credentials stay on the server and the access token is cached between requests. The browser only ever talks to the app’s own endpoints.
  • Built around the credit budgetEvery call is metered against the daily credit allowance, rate-limit replies are honored, and if the API goes down the map shows the last good data instead of breaking.

From API call to the Denver hub board

  1. OAuth2The server signs in to the OpenSky API
  2. Live positionsOne poll, cached and shared by every tab
  3. Denver hubDepartures, arrivals and ground traffic sorted
  4. Flight recordsRoutes looked up and labelled as inferred