How solid are the numbers on this page?

They are labelled by kind. Verifiable figures trace to a report or a commit: crawl counts, backlink audits, clicks, authority scores, running costs. Measured figures are a straight account of how long the work took, before and after. Comparison figures, like a 6-figure analyst salary or an $8,000 to $20,000 rebuild quote, are market rates rather than invoices.

When is Two Fold TT not the right call?

If you already know exactly what to build and just need hands, a developer is cheaper. If the problem is that nobody owns the process, software will not fix it.

How does an engagement start?

With a 30 minute call at no charge. Every engagement then starts with a diagnosis rather than a build, and scope and fixed pricing follow once the work is understood. If automation will not help, that is the finding.

What is needed from the client to start?

Read access to your analytics, ad accounts and site, one 60 minute kickoff, and one person to go to with questions. Build work after the diagnostic is quoted as fixed blocks off the roadmap, each approved before it starts. No retainer is required to start.

Why do most AI initiatives fail?

Rarely because the technology failed. S&P Global Market Intelligence found 42% of businesses scrapped most of their AI initiatives in 2025, and MIT Project NANDA's preliminary GenAI Divide report put the share of enterprise pilots with no measurable P&L return at about 95%. The common cause is building the right thing for the wrong problem, and the mistake is usually not obvious for months. That is why the work here starts with a diagnosis rather than a build.

Two Fold TT — Marketing operations and workflow automation, Melbourne

The thing nobody puts on their website

42% of businesses scrapped most of their AI initiatives last year.

Rarely because the technology failed. Usually because the mistake wasn't obvious until four months and forty thousand dollars later.

S&P Global Market Intelligence, 2025 (opens in a new tab), survey of 1,000+ businesses in North America and Europe. Up from 17% the year before.

95% of enterprise AI pilots returned nothing measurable.

Thirty to forty billion dollars spent, and almost none of it reached the accounts. The report is preliminary and not peer reviewed, so weigh the exact figure accordingly. The direction is not in dispute.

MIT Project NANDA, The GenAI Divide, 2025 (opens in a new tab), 150 leader interviews, 350 employees surveyed, 300 public deployments analysed.

Statement 1 of 2, S&P Global Market Intelligence

So the work starts with proof, not a proposal.

What gets built

Four kinds of work. Most clients only need one.

Whichever is costing you the most time or money right now is where we start.

Intelligence pipelines

Know what is moving in your market before your competitors do, without paying someone to read the internet all morning.

One runs for 28 cents a day, against six figures for a dedicated analyst. It tells you the moment it stops working.

Analysis and audit agents

Catch the expensive mistake while it is still cheap.

Pricing errors before a customer sees them, website errors before Google penalises you for them, ad spend wasted before the budget is gone.

Marketing production agents

Ads, email, content and reporting produced at volume, checked against your brand guide before a person ever sees them.

Your team's time goes to judgement calls, not proofreading.

Workflow automation

The repetitive task that eats a day a week, handled on its own.

You can step in any time, and anything unclear gets flagged for you instead of guessed at.

Built on, integrated with, or migrated between

Claude · Amazon Bedrock · AWS · n8n · Docker · GA4 · Search Console · Google Ads · Shopify · Semrush · Screaming Frog · OpenAI · HubSpot · Adobe · Figma · Databricks

Selected work (what we deliver)

Six builds. Every one started with a wrong assumption.

National law firm · 30+ offices · content operations

Came in with

Every brief took close to 2 weeks.

The bottleneck wasn't the writing. It was starting from a blank page every time.

The delay came from handovers, collaboration gaps, and no consistent structure to start from. Humans still do the research and set the approach.

AI drafts each brief to 87 to 95 percent complete, then runs its own cross-reference and QA checks to keep output consistent over time.

A copywriter refines it and a lawyer signs off on every reference before it goes out.

2 days

Measured. Per brief, down from 2 weeks.

Online camera retailer · €2.1M organic

Came in with

Products weren't showing up in Google.

Their agency said nothing was wrong.
Eight million junk web addresses said otherwise.

An AI-assisted crawl audit found the cause in a day, work that manual analysis would have taken 2 weeks to uncover: 8 million junk web addresses burying 67,000 real pages, and a test version of the site quietly live and taking real orders.

They came for a fix and left with a twelve month roadmap.

1 day

Verifiable. Of analysis, down from 2 weeks manual.

Tour operator · 100+ reviews · gap analysis

Came in with

Their search authority score sat at 2.

They assumed they'd been penalised.
They'd been targeted, by a network they'd never paid.

Nobody could explain why either. 98 of 111 backlinks turned out to be spam, planted by a network that had latched onto the site without anyone's knowledge. Inherited, not self-inflicted, which moved the fix from urgent to routine.

The client had already spent on accreditation memberships without knowing which ones actually helped. Our AI-driven competitive analysis found the real answer: a rival's stronger ranking traced back to a single accreditation backlink, built to Google's schema standards. We pointed them to the same one. Their ranking moved.

A custom AI integration, built on MCP and connected to Analytics, Search Console, and their heat-mapping tool, now keeps the profile monitored, not just fixed once.

$295

Verifiable. A year to close the real gap.

Mobile bike servicing · 5.0★, 169 Google reviews

Came in with

A booking form that could actually quote.

Quoted as two weeks of developer time. The build wasn't the expensive part, scoping it was.

Service cost, service area, delivery fee, and day availability, all worked out automatically. That's normally two weeks of developer time plus a business analyst to scope it. AI handled the build, a human directed each feature, done in 7.5 hours.

While reviewing the code, we found a pricing reference showing $900 where it should have read $90, buried in backend logic no customer would ever see. AI now runs routine checks on the code itself, and any pricing change triggers a double confirmation before it goes live.

7.5 hrs

Measured. To build, down from a 2 week quote.

Client names are published only with individual permission, which is why some of the work above is described by sector rather than by name. On a call I can walk you through any of it in detail.

How this actually works

Diagnose. Build. Run. Iterate.

The expensive failure is rarely a bad build. It is building the right thing for the wrong problem.

Diagnose

Primary data before opinion. Your analytics, your site, your actual records, not your brief. Anything that contradicts what you expected gets flagged, not smoothed over.

Findings ranked by severity, effort and who fixes it
Build

Shipped in stages against a written requirement, then checked on the real, live site. Never just confirmed in a build log, because a clean deploy can still leave the old version showing to a real visitor.

A working system, a rollback point, a changelog
Run

Monitored on a schedule, not checked only when something breaks. Cost per run is tracked from day one, and if anything drifts outside its normal range, it raises a flag before a client notices.

Health checks, running costs, an off switch, a handover pack
Iterate

A system built for last quarter's business is already stale. Each cycle re-checks assumptions against what's actually true now, and the build gets adjusted, not patched over.

Assumptions re-checked against what is true now

Every cycle loops back into Diagnose

Where to start

Start with the diagnosis, not the build.

Bring the process that is costing you the most. Thirty minutes is usually enough to tell you what is actually wrong, what it would take to fix, and whether I am the right person to do it.

The 30 minute diagnostic Free

You leave knowing three things: what is actually broken, what it would take to fix, and what it would cost. No deck, no pitch. Scope and fixed pricing follow the call, once I know what the work actually is.

Book a diagnostic call

Evidence

Every number on this page, labelled.

Every figure on this page is one of three kinds. Here is how solid each one is, so you can weigh them yourself.

Verifiable

Traceable to a report or a commit

Crawl counts, backlink audits, clicks, authority scores, running costs. Just ask and I will show you where the number came from.

Measured

My own time, before and after

A straight account of how long the work took, timed before and after. No projections, no modelled savings.

Comparison

What the alternative would have cost

A 6-figure analyst salary, an $8,000 to $20,000 rebuild quote. Market rates, not invoices. Treat them as the ballpark they are.

When I am not the right call

If you already know exactly what to build and just need hands, a developer is cheaper than me.
If the problem is that nobody owns the process, software will not fix it, and I will say so on the call.

Don't take my word for it.

My job history, current role included, is public on LinkedIn.

View LinkedIn profile (opens in a new tab)

Thirty minutes. If I'm not right for this, I'll say so and point you to who is.

Book a 30 minute call

I find what is actually broken, then build the fix.

42% of businesses scrapped most of their AI initiatives last year.

95% of enterprise AI pilots returned nothing measurable.

Four kinds of work. Most clients only need one.

Intelligence pipelines

Analysis and audit agents

Marketing production agents

Workflow automation

Built on, integrated with, or migrated between

Six builds. Every one started with a wrong assumption.

The bottleneck wasn't the writing. It was starting from a blank page every time.

A senior person started every day with 45 minutes of manual scanning.

Their agency said nothing was wrong.
Eight million junk web addresses said otherwise.

They assumed they'd been penalised.
They'd been targeted, by a network they'd never paid.

Quoted as two weeks of developer time. The build wasn't the expensive part, scoping it was.

A spam filter would have fixed it once. Bots change their approach weekly.

Diagnose. Build. Run. Iterate.

Diagnose

Build

Run

Iterate

Start with the diagnosis, not the build.

The 30 minute diagnostic Free

Every number on this page, labelled.

Traceable to a report or a commit

My own time, before and after

What the alternative would have cost

When I am not the right call

Don't take my word for it.

Bring your messiest process.

42% of businesses scrapped most of their AI initiatives last year.

Four kinds of work. Most clients only need one.

Intelligence pipelines

Analysis and audit agents

Marketing production agents

Workflow automation

Built on, integrated with, or migrated between

Six builds. Every one started with a wrong assumption.

The bottleneck wasn't the writing. It was starting from a blank page every time.

A senior person started every day with 45 minutes of manual scanning.

Their agency said nothing was wrong.Eight million junk web addresses said otherwise.

They assumed they'd been penalised.They'd been targeted, by a network they'd never paid.

Quoted as two weeks of developer time. The build wasn't the expensive part, scoping it was.

A spam filter would have fixed it once. Bots change their approach weekly.

Diagnose. Build. Run. Iterate.

01Diagnose

02Build

03Run

04Iterate

Start with the diagnosis, not the build.

The 30 minute diagnostic Free

Every number on this page, labelled.

Traceable to a report or a commit

My own time, before and after

What the alternative would have cost

When I am not the right call

Don't take my word for it.

Bring your messiest process.

Their agency said nothing was wrong.
Eight million junk web addresses said otherwise.

They assumed they'd been penalised.
They'd been targeted, by a network they'd never paid.

Diagnose

Build

Run

Iterate