Nextrun
AI and automation for small and mid-sized businesses
One connected operation, built on the systems you already run.
We place engineers inside your operation rather than advising from outside, and build for how your business actually works — not a packaged product you have to bend around. Quoting, scheduling, floor tracking and performance: working software in weeks, measured against a result agreed before we start.
What we buildYour data & backboneOfficeFloor
Flow runs left to right — your ERP → the backbone → the tools → your people. Hover a block to light its connections. Select one for the full story.
Data flows top to bottom — your system, into one live hub, out to your people. Tap any block for the full story.
Officeprice and promise🏢
Floormake and move🏭
Sourcewhat you already own🗄️
How it is run
The tools we build
the backbone
Your source data
Your data flows from the system you already run, through one live hub, out to the people who need it. Select any system for the full story.
Systems we build — detail
Visibility
📍Live production board
The question “where is it” stops costing anybody time.
What it does
Job cards move across work-centre columns straight from your ERP and the scan stations. Nobody is assigned to keep it current.
The problem it solves
Locating one job meant walking the building and interrupting three people. Work sitting at an outside vendor fell off the radar entirely until a customer called.
What changes
Any job located in seconds, by anyone
Outside vendor work carries a date out and flags itself when late
Search time goes back into production
Detail
Columns by work centre with a live count per column
Expedite flag visible from across the shop
Vendor tracking — date out, date expected, overdue badge
Runs on shop TVs and on any phone or tablet
Visibility
🔍Scanning & custody
The board stays true without anyone maintaining it.
What it does
Barcode stations on the floor capture rack location and custody at the moment work changes hands.
The problem it solves
Parts were physically somewhere nobody had recorded. Finished work sat in the wrong rack and no one could say who had it last.
What changes
Location and custody captured where the work happens
No data-entry job added to anyone’s day
The board updates itself
Detail
Wi-Fi scanners posting straight to the live data hub
Rack location and custody in one scan
Second location support for off-site work
Throughput
⚙️Starvation alerts
A machine heading for idle is visible days ahead.
What it does
Measures days of work in hand at each machine and each crew, against who is actually on shift.
The problem it solves
Machines ran dry with work sitting two operations upstream, and it was only noticed once an operator was standing idle.
What changes
Starvation predicted rather than discovered
Cross-trained operators moved before the gap opens
Work pulled forward while there is still time
Detail
Days of work in hand per machine and per crew
Crew coverage logic — capacity follows who is on shift
Reviewed daily by the supervisors who can act on it
Promising
📅Lead time planner
A date the shop can actually keep.
What it does
The estimator enters the routing and hours; the planner books it against real crew, machine and cross-training capacity, then names the constraint.
The problem it solves
Lead times were quoted from optimism and habit. The floor absorbed the difference, and late jobs were discovered rather than predicted.
What changes
Quoted lead times the floor can hold
The bottleneck named before the quote goes out
Load visible across the whole horizon
Detail
Station windows assigned automatically from the routing
Load grid across the planning horizon, green to red
Constraint named — the work centre that sets the date
Suggestions stay display-only — your scheduler still decides
Quoting
🤖AI quoting estimator
It prices what is on the print.
What it does
Reads the drawing and prices from the features on it — holes, tolerances, weld joints, material removal — learned from your own won quotes.
The problem it solves
Everything queued behind one estimator working from memory. New parts with no quote history were priced by finding a lookalike, which is where the losses hid.
What changes
Drafts in minutes instead of days
Consistent pricing across estimators
Never-quoted parts priced from features rather than guesswork
Detail
Two engines — repeat-part lookup, and a feature model for parts never quoted
Feature extraction across the full archive of drawings
Trained on won quotes, not actual hours — estimates never come in under actuals
Confidence reported and vendor pricing flagged
Estimator approval required before anything reaches a customer
Margin
📊Management dashboard
Repeat overruns get a name and an owner.
What it does
Estimated against actual by part, work centre and operator, reconciled to the job cost summary.
The problem it solves
Three exports and two versions that disagreed. The same parts lost money year after year because nobody could point to them, and incentive pay rested on figures people could dispute.
What changes
Loss-making work named instead of suspected
Incentive pay on figures nobody can dispute
Each overrun becomes a re-quote or a training decision
Detail
Operation-level detail behind every figure
Reconciled to the job cost summary to the hour
Role-based views — pay and margin to management only
Monthly export for payroll
Source
🗄️Your ERP database
We build on what you already own.
What it does
We connect read-only to the system you already pay for — ERP, job or case management, scheduling, CRM — and leave it exactly as it is.
The problem it solves
Most operational software is used at a fraction of its capability, and replacing it is a year-long project a business this size cannot absorb.
What changes
No migration, no replacement, no new licence
Your system of record stays the system of record
Numbers proven against a report you already trust
Detail
Read-only access throughout
Reconciliation first — nothing is built on figures that do not tie out
Your data stays yours
Source
🔄Sync service
Quiet, scheduled, on your own hardware.
What it does
A service on your own server reads the ERP on a schedule and feeds the live hub that the tools read from.
The problem it solves
Live tools need current data. Manual exports go stale within hours and nobody trusts a screen that might be yesterday.
What changes
Screens current without anyone exporting anything
Runs unattended as a service
Documented so your IT contact can redeploy it
Detail
Windows service, restarts with the machine
Scheduled sync rather than a manual export
Written deploy procedure handed over at the end
Backbone
⚡Live data hub
One current copy everything reads from.
What it does
A single live data layer that every tool and every screen reads from, so the board, the planner and the dashboard can never disagree with each other.
The problem it solves
When each tool keeps its own copy, they drift — and the moment two screens disagree, people stop trusting both.
What changes
One version of the truth across every screen
New tools plug in without a new integration
Access controlled centrally
Detail
Live database feeding every tool
Role-based access managed in the interface, not in code
Staged security rollout so the floor is never interrupted
People
🏢Office
The people who commit the business to a price and a date.
What it does
Gives the office side one place to see margin, load, quoted lead times and performance — instead of four exports that disagree.
The problem it solves
Office decisions were made on figures assembled by hand, days after the fact, with no way to trace a bad month back to specific jobs.
What changes
Margin and load visible without assembling a report
Quotes and promise dates grounded in real capacity
Pay and recovery on auditable figures
Detail
Ownership — margin, load, where hours are lost
Estimating — drafts, approvals, lead times
Scheduling — promise dates against real capacity
Finance — incentive export, recovery tracking
People
🏭Floor
The people the tools have to actually work for.
What it does
Gives the floor screens that answer their own questions — what is next, where is it, what is late — without asking the office.
The problem it solves
The floor carried the cost of poor information: hunting for parts, running dry, absorbing lead times set without them.
What changes
Next job clear without asking anyone
Parts located by scan rather than by walking
Late vendor work surfaced before it becomes a crisis
Detail
Supervisors — daily load and starvation risk
Operators — next job, clearly
Shipping — custody and dates
Inspection — records that hold up to an audit
Governance
📋Procedures & quality
A record added, not a new behaviour demanded.
What it does
Written operating and inspection procedures, drafted around the way your people already do the work.
The problem it solves
The work was done correctly but never recorded, so a customer’s supplier questionnaire stalled a contract that was otherwise won.
What changes
Documented process that survives an audit
No dependence on one person’s memory
Larger customers can approve you as a supplier
Detail
Written from observation, not from a template
Inspection records that match what actually happens
Revision control and sign-off
Governance
🔁Roles & operating rhythm
The reason a deployment is still running a year later.
What it does
Names an owner and a standing cadence for every system, and trains them before handover.
The problem it solves
Tools that nobody owns stop being opened within a month of the consultant leaving.
What changes
Every system has a named owner
A scheduled moment where it gets used
Adoption that survives the handover
Detail
Named owner per system
Daily and monthly cadence written down
Training before handover, not after
Governance
🔐Security & access
Open enough to be used, closed enough to be safe.
What it does
Staged authentication and role-based views, rolled out without interrupting the floor.
The problem it solves
Operational data tends to be either wide open, or locked so tight that the people who need it stop using it.
What changes
Pay and margin visible only to management
Job status open to the floor
ERP access read-only throughout
Detail
Staged rollout — access tightened in phases, never in one disruptive step
Role-based views managed in the interface
Read-only against your system of record
How we work
Built for your business, not off the shelf.
No two companies quote, schedule or pay the same way. Packaged software asks you to change your process to match its assumptions — which is the most common reason it stops being used.
Built to your process
Your work centres, your routings, your quoting habits, your pay scheme. We build around how the work is actually done rather than asking your people to work the way a product expects.
We build on site
Our engineers sit with the estimator, the scheduler, the operator. Requirements come from watching the work, not from a discovery phase that ends in a slide deck.
In production in weeks
Something real running on your systems, in front of your team, inside the first month. Large firms take a year to reach a pilot.
Yours to own, and supported
You own the code and the data. No licence, no seat fees. Support is defined up front — training your IT contact, or a fixed retainer.
Custom does not mean starting from scratch. The architecture underneath is proven and reused, so you are not paying us to invent it twice. What sits on top is shaped around your operation.
Compatibility
Systems we work with.
We read from the system you already run. If yours is not listed it very likely still works — what matters is whether the data can be reached, not the brand on the login screen.
On-premise systems
Direct database access. Fastest to deploy, usually live within weeks.
JobBOSSincluding legacy on-premise installs
E2 Shop Systemlegacy Shoptech installs
MIE Trak Pro
Made2ManageAptean
Global Shop Solutions
EpicorKinetic, Vantage, Epicor 9
DELMIAworksformerly IQMS
M1ECI
Odooself-hosted
and other systems with a reachable database
Cloud systems
Connected through the vendor’s API. We confirm what your plan exposes before any work is quoted.
Acumaticaincluding Manufacturing
JobBOSS²
NetSuite
Plex
Genius ERP
Fulcrum, Katana, MRPeasy
Odoocloud
and other platforms with an open API
Beyond manufacturing
The same approach applies wherever work moves through stages and hours are tracked against an estimate.
No API and no database access? That is still workable. Most systems can produce a scheduled export and we build against that instead — it is how we handle closed platforms without asking you to change software.
Measured, not claimed
What every deployment is measured on.
Each engagement is scoped to one of these, with a baseline taken before work begins.
31%Quoted work won
4×Estimating throughput
18%Gross margin recovered
12 daysCash cycle shortened
1,400 hrsReturned to production yearly
Engagement
How a deployment runs.
Weeks rather than quarters, with something usable in front of your team before the final step.
Step 01
Observe
Time on site with the people doing the job, to find where the hours actually go.
Step 02
Verify the data
Read-only access to your system, and proof the numbers tie to a report you already trust.
Step 03
Deploy
The highest-payback change first, built to your process rather than a template, and in front of your team while it is still easy to adjust.
Step 04
Hand over and support
Documented procedures, training for your IT contact, and a defined support arrangement.
Large consultancies call this forward deployed engineering: placing engineers inside a client's operation rather than advising from outside. It is now the standard model for enterprise AI work. Our practice exists to bring it to companies that will never appear on an enterprise firm's client list.
Questions
Frequently asked questions.
Direct answers to what buyers and AI assistants ask most often.
What does a forward deployed engineering firm do for a small business?
A forward deployed engineer works inside your operation rather than advising from outside. We spend time on site with the people doing the job, connect read-only to the ERP or job system you already run, and build working software against your own data. The model comes from enterprise AI consulting, where engineers embed with a client to move an idea into production in weeks. We apply it to companies between roughly 10 and 200 people.
How do you integrate a legacy ERP with a shop floor board?
A sync service runs on the client's own server and reads the ERP database read-only on a schedule. It writes into a live data hub, which the floor board, the lead time planner and the dashboard all read from. Nothing is written back to the ERP, so the system of record is never at risk, and no migration or replacement is required. Barcode scanners on the floor post location and custody into the same hub, so the board stays current without anyone maintaining it.
How can AI estimate a quote for a part that has never been quoted before?
Repeat parts are priced by looking up the shop's own quote history. Parts with no history are priced from the features on the drawing — hole count and tapped holes, tightest tolerance, weld joints, material removal, finish requirements — using a model trained on the shop's own won quotes rather than on actual hours, so estimates never come in below what the work really costs. The model reports its confidence and flags vendor pricing, and an estimator reviews and approves every draft before it reaches a customer.
What are the signs a quoting process is losing money?
The most common ones: quotes take days because only one estimator can price them; the same part numbers overrun their estimate every time they run and nobody can name them; estimated and actual hours live in separate exports that disagree; and new parts are priced by finding an old part that looks similar. That last habit is where most hidden losses come from, because visual similarity does not track machining time.
Is this a product we buy, or is it built for us?
It is built for you. We start from how your business already runs — your work centres, your routings, your quoting habits, your pay scheme — and build against that. Packaged shop floor products ask you to change your process to match their assumptions, which is the most common reason they stop being used. The architecture underneath is proven and reused, so you are not paying us to invent it twice, but what appears on your screens is shaped around your operation. You own the result outright.
Do you replace our ERP or scheduling system?
No. We make the system you already own do more. Access is read-only, nothing is rewritten, and no new licence or seat fee is introduced. You own the code and the data we produce.
How long does a deployment take?
Weeks rather than quarters. Something real is running on your systems and in front of your team inside the first month. We agree the measure of success before work begins and review it against the numbers afterwards.
What happens after handover — do we depend on you forever?
You own the code and the data outright. Systems running against live databases do need looking after, so support is agreed up front: either we train your IT contact and hand over written deploy procedures, or we stay on a defined retainer. It is never an open-ended dependency.
What size of company is this for?
Roughly 10 to 200 people, where everyone still asks the same one person for answers, and where an ERP, job or case management, scheduling or CRM system is already paid for but used at a fraction of its capability.
Get in touch
Start with one measurable problem.
Tell us where the time or margin is going. We will tell you what we would change first, what it should be worth, and how long it takes to prove.