Most AI consultants advise.I build the systems.
I'm James Webb. I founded LINK in 2012 and grew it into a nine-business group, and I've spent that time building the systems and processes those businesses run on - across accounting, marketing, property, finance, recruitment and creative. Different industries, the same problem underneath: work that should be automated, data that should talk to itself, and decisions that should be made on something better than a hunch.
Most of that work now starts in the same place: your numbers are in Xero, the rest of the business is in four other systems, and nobody can see the whole picture in one place. Connecting those and turning them into a dashboard that actually explains the business is the bulk of what I get asked for - and because LINK is an accounting group, we can read the panel as well as build it.
James Webb
Founder, LINK Group
6
industries built for: accounting, marketing, property, finance, recruitment, creative
9
businesses running on systems built in-house
$40k
a year of software subscriptions replaced by building it
1
person you deal with, start to finish
Three ways in. Buy any of them on their own.
Connect it, then show me the numbers
Xero joined to the systems the rest of the business runs on - job management, CRM, payroll, POS, project tooling - and the result presented as one panel that answers how the business is actually going. We implement the right tools rather than rebuilding what already exists: Xero's own reporting, Fathom, Power BI, whatever fits, and custom work only where off-the-shelf stops. Then LINK's accountants help you read it, which is the half software cannot sell you.
Your numbers are in five places and none of them agree.
Dashboards & management reporting →Build the system
Custom AI backends, internal tools, CRMs, lead engines, scoring and routing, document and data pipelines. Built on a real database with your process as the model, not an off-the-shelf product bent until it nearly fits. Where an open-source model is the right answer it gets deployed and run properly rather than rented by the token.
You know the problem and want something that works.
Strategy and roadmap
Where AI genuinely fits in your business and where it does not, what to automate first, what to leave alone, build versus buy, what it costs to run, and the governance around anything customer-facing. Delivered as a roadmap you could hand to another developer, not a deck.
You know AI matters and want an honest map before spending.
It costs less than it used to, and you own it at the end.
Custom software has a reputation earned in a different decade - six figures, a year, and a specification argued over by people who will not use it. That is not what building looks like now. The tooling has changed what a small, experienced team can ship, and the work that used to justify an enterprise budget frequently does not any more.
The honest version: it is cheaper than you have been quoted, and it is not free. What it replaces is usually a stack of subscriptions charged per seat forever - across LINK's own businesses that came to about $40,000 a year, which is the number in the band at the top of this page.
And what gets built is yours. The code, the database, the logic of how your business actually works - handed over, not licensed back to you. No per-seat fee that grows as you hire, no vendor holding your data hostage at renewal, and no product roadmap you do not control deciding what your business can do next year.
Where an off-the-shelf tool genuinely does the job - Xero's own reporting, Fathom, Power BI - we will tell you to buy it rather than sell you a worse version of it. The custom work is the connective part between your systems, which is the bit nobody sells because it only exists in your business. That is the part worth owning.
The credential is the software. Here is what it is.
Everything below is built and working. Most of it is running a LINK business right now, several are public and you can go and use them before you talk to me, and the two marked pre-rollout are finished but not yet carrying production traffic. That distinction matters: a page like this is only worth anything if it is accurate.
Custom AI models, built not bought
Purpose-built models running the group's core operations: CRM, lead management, automation, team management and customer management. Not a wrapper around someone else's product with a prompt in front of it - models built for these specific jobs, on data structured for them. This is the part most AI consultants are describing when they say 'we can help with AI', and the part very few have actually done end to end.
It replaced $40,000 a year of software
The CRM, lead management and automation layers running LINK were SaaS subscriptions before they were built here - around $40,000 a year of them across nine businesses. Building removed that, and removed something worth more than the money: the ceiling. The system now matches how the group actually works, instead of the group bending to fit what the product allowed.
rank.™ and rankbot
The automated SEO engine behind LINK Marketing's productised plans. It runs keyword tracking, on-page fixes, content scheduling and competitor monitoring unattended, which is the reason productised SEO starts at $250 a month there and $1,400 elsewhere in the market. A machine doing the repetitive hours so strategists spend theirs on judgement.
InstantScan
A server-side scorer that fetches any URL, runs ten weighted on-page checks and returns a result out of 100 in about a minute. Nothing is gated, and every run still posts the domain and its score into the sales channel - a lead signal without asking the visitor for anything. The same function then runs paying clients' monthly health sweeps on a cron: one code path, so a prospect's free scan and a client's monthly report are literally the same measurement. The published criteria are rendered straight from the check catalogue too, so what the page promises cannot drift from what the software does.
The Rescue assessment engine
A six-question scoring tool for directors under financial pressure that bands into Green/Amber/Red with a fast lane for director penalty notices. The score is recomputed server-side on submission so it cannot be gamed from the browser, and every lead delivers three ways at once - database record, Slack with urgency markers, and email.
The nine-property estate
Nine LINK businesses rebuilt onto one modern stack, each with its own lead capture, attribution (UTM and gclid on every enquiry), structured data and reporting - migrated off legacy WordPress without losing the rankings that paid for them.
Estimators, calculators and qualifiers
Fee and quoting estimators, a business valuation model, borrowing power and retirement calculators, a property appraisal wizard, a brand health check, an eight-domain business performance diagnostic. Built for six different industries, and each one is a lead engine wearing a useful tool as a coat.
Lead routing and delivery
The plumbing under all of it: enquiries scored, routed to the right division, delivered to the right desk, and attributed back to the campaign that produced them. Including the arrangement that sends this page's enquiries straight to James rather than into a queue.
The group lead platform
A central system taking every enquiry across nine businesses, applying routing rules per division and per desk, and holding the record. Built and working, in final setup ahead of rollout - which is the honest status rather than a claim it is already carrying production traffic.
Built · pre-rollout
The client portal
Client-facing reporting and account access built to sit over the same data, so clients see the system's own numbers rather than a screenshot pasted into a slide. Built and functional, running in demo ahead of release.
Built · pre-rollout
And not just for us.
Three client builds, described without naming them. Different problems, one thread: no product on the market fitted how the business actually ran. Sometimes that is a gap, and just as often it is the opposite - a platform doing a great deal you never asked for while missing the one thing you needed.
Commercial tiling contractor
The problem
Not a business that had outgrown its software - a business the software never fitted. They were on simPRO, which is a capable platform, and that was part of the problem: it did a great deal they had no use for, and the things they did need it never did the way they wanted. Every general-purpose trades product is built for the average of its market, and nobody is the average.
What was built
One custom system modelling how this business actually runs - staff, jobs, scheduling and quoting on a single data model, with the features they use and none of the ones they do not.
They moved off simPRO onto it, and saved $55,000 in systems cost.
The saving was the smaller half. What mattered more was the revenue that had been leaking through project management the old setup could not keep up with - jobs that slipped, work that went unbilled, margin lost between the quote and the invoice. That is the real cost of a system that does not fit, and it never shows up on a subscription line.
Large online retailer
The problem
Out-of-the-box inventory and the standard web integration between it and the store were not good enough for how the business actually trades. Stock, orders and customers each lived in their own product, and the joins between them were the weakest part of the operation.
What was built
Their own back end: inventory, customer management, and the web integration tying both to the storefront - one system rather than three products and the glue between them.
Stock, customers and the store now run on one back end instead of three.
Multi-service online business
The problem
Several different services sold online, and the entire operation run out of an email inbox. A job arrived as an email, someone worked it, then moved to the next. No record beyond the thread it came in on, no view of who was carrying what, and no system on the market that would bring the whole thing together.
What was built
One back end covering CRM, project management and people management. Leads and jobs land, convert, and route to the right person to process, with workload per person visible so capacity is managed rather than guessed at. Email and follow-ups run on automation, abandoned carts get chased, and customers get cross-referred between services instead of being sold one thing and forgotten.
The inbox stopped being the system of record.
This one was never a cost story, and it is worth saying so. Nothing off the shelf existed that would run their services, their follow-ups and their carts as one thing, so the return is capacity rather than savings: the same team absorbing more work, a business that can take on volume without hiring to cover the admin, and an operation that no longer depends on someone remembering. Efficiency and headroom are harder to put a number against than a cancelled subscription. They are usually worth more.
Three things an AI consultant should tell you and usually doesn't.
Most businesses do not need a custom AI system
They need their existing tools connected properly and two or three processes automated. That is a smaller, cheaper job and you will be told so in the first conversation. The market is full of people who will sell you a build regardless.
AI is not the interesting part
The interesting part is the data model underneath it. An AI layer over messy, badly-structured data produces confident nonsense faster. Most of the work in a good build is the boring part, and skipping it is why so many AI projects quietly get switched off.
A human approves anything customer-facing
Content, replies, quotes, anything a customer sees. Not because the models are bad, but because your name is on the output and nobody outsources their reputation to a probability distribution.
It starts with a technical discovery.
Step one
The conversation
You explain the problem. You get a straight read on whether building something is justified, roughly what it involves, and what the cheaper alternatives are. Free, and genuinely no obligation.
Step two
Technical discovery
Paid, scoped, and it produces a specification you own: the data model, the integrations, the build sequence and the cost. Take it elsewhere if you want - it is yours either way.
Step three
Build, in the open
A working version early so you can steer it, then iteration against real use. No six-month silence ending in a reveal.
Here for AI in your SEO and marketing?
That is a different job to building a system, and it is already productised. LINK Marketing runs it on the tooling described above - so you get the same engine without commissioning a build.
rank.™ — automated SEO
The engine itself, sold as a plan. Keyword tracking, content, on-page fixes and competitor monitoring run by the machine, directed by a strategist. From $250 a month, no lock-in.
See rank.™ →
AI search visibility
Being cited by AI assistants and AI Overviews, not just ranked by Google. Structured data, content built to be quoted, and the measurement to know whether it is working.
How LINK runs SEO →
Questions, answered on the page.
Can you connect Xero to the rest of our software?
That is most of the work. Xero holds the financial truth and almost nothing else - the jobs are in job management, the pipeline is in a CRM, the hours are in payroll or a rostering tool, and the picture only exists if someone exports all of it into a spreadsheet every month. Connecting those so the numbers arrive on their own, and land somewhere that explains what they mean, is the job. Xero's API is good and most of the common systems have one; where a system genuinely does not, we say so before you spend anything.
Should we just use Fathom, or Xero's own reporting?
Often, yes - and we will tell you so. If your question is 'how is the business tracking against last year' and everything you need is already in Xero, then Xero's own reporting or Fathom answers it for a subscription, and building something custom would be selling you a worse version of a product you already pay for. Xero bought Syft in 2024 and is folding that analytics layer in natively, so this only gets more true. The point where those tools stop is when the answer needs data that is NOT in Xero - jobs, utilisation, pipeline, stock - and that is where building is worth it.
What does a custom dashboard or system actually cost?
Less than you have probably been quoted, and it is worth saying plainly because the reputation of custom software was earned in a different decade. The tooling has changed what a small experienced team can ship, and work that used to justify a six-figure enterprise budget frequently does not now. The honest framing is not 'cheap' - it is that the comparison is usually a stack of per-seat subscriptions charged forever. Across LINK's own businesses, building instead of renting took about $40,000 a year out of the software bill.
Do we own what you build?
Yes - the code, the database and the logic of how your business works. It is handed over, not licensed back to you. No per-seat fee that grows every time you hire, nobody holding your data at renewal time, and no product roadmap you do not control deciding what your business is allowed to do next year. That is the substantive difference between commissioning something and subscribing to something, and it is the reason the sums work out differently over three years than they look in month one.
What does AI consulting actually mean here?
Three things, and you can buy any of them on their own. Building systems - custom AI backends, CRMs, lead engines, internal tools, automation. Strategy - an honest roadmap of where AI fits in your business and where it does not. And growth - using AI to be found in AI search and to run content and follow-up at a scale you could not staff. What it does not mean is a workshop, a maturity assessment or a slide deck.
What makes you different from every other AI consultant right now?
Most are reselling somebody else's tool or advising on work they have never done. LINK's own nine businesses run on systems built in-house across six industries - an automated SEO engine, custom models behind the CRM and lead management, scoring and routing, a stack of calculators and lead engines - and you can go and use several of them right now. Outside the group: a tiling contractor moved a whole trades operation off simPRO and took $55,000 out of their systems cost, a retailer runs inventory, customers and their storefront on one back end instead of three products, and a services business stopped running itself out of an email inbox. That is a different conversation from a capability statement.
Do you build custom models, or wire up someone else's product?
Custom. The models running LINK's CRM, lead management, automation, team management and customer management were built for those specific jobs rather than bought and configured. That is the difference between a system that fits your process and one your process has to bend around. Where a hosted model genuinely is the right tool for part of a build it gets used, and the cost, privacy and control trade-off gets made with you rather than for you - but the architecture is designed, not assembled from whatever the vendor allows.
Can this replace software we currently pay for?
Often, and that is frequently what pays for the build. LINK's own CRM, lead management and automation layers were subscriptions before they were built in-house - about $40,000 a year of them across nine businesses - and replacing them removed that cost along with the ceiling those products imposed. Whether it stacks up for you is a maths question, and the discovery answers it honestly: if your current stack is cheaper than replacing it, you will be told that.
How much does it cost?
No published price, because a two-week integration and a six-month platform build are not the same product and pretending otherwise wastes your time. It starts with a technical discovery: what you run today, where the friction actually is, and whether a build is justified. You get a scoped quote and a written recommendation out of that, including the recommendation to do nothing if that is the right call.
How long does it take?
Connecting existing systems and automating a process or two is usually weeks. A custom platform with its own data model is months, because the discovery and modelling are the job and rushing them means building it twice. You get the timeline before you commit, and the first working version is deliberately early so you can steer it.
Will my data be safe?
It is the first thing scoped, not the last. Where data cannot leave your infrastructure, the system is built so it does not - which is one of the strongest arguments for a self-hosted open-source model. Where hosted models are used, what goes to them and what never does is decided explicitly and written down.
Can you work with our existing developer or IT team?
Yes, and often that is the cheapest way to do it: the strategy and architecture come from here, your team builds it, and the roadmap is written to be handed over. Where there is no internal team, the build runs here end to end.
Do you work outside Brisbane?
Yes. We are in Fortitude Valley and most of this work is remote regardless of where the client sits, so the search runs Australia-wide. Brisbane clients get in-person discovery where it is useful.
Is AI going to replace my SEO or my marketing team?
No, and anyone telling you otherwise is selling something. What it changes is the leverage: the repetitive hours get done by a machine and the people spend theirs on the decisions. That is exactly how rank.™ works and why productised SEO can start at $250 a month - not because a person was removed, but because their time moved to where it was worth more.
What happens in the first conversation?
You explain the problem. You get a straight read on whether it is worth building something, what it would roughly involve, and what the alternatives are - including the cheaper ones. If there is a build worth scoping, the next step is a paid technical discovery that produces a specification you own, whether or not you build it here.
Tell me what's not working.
This form goes straight to my inbox - not a sales team, not a queue. If the honest answer is that you do not need a custom build, that is the answer you will get, and it will save us both a month.
Not sure whether you need a build or a plan? Say so in the message and you will get a straight answer, including when the answer is rank.™ at $250 a month rather than anything custom.