An AI that reads the tax alerts before the team does
GIC’s tax team was spending a tenth of its time reading advisory email to catch changes in tax law. A model now categorises each item first, checked against the team’s own judgement.
The problem
Tax law changes in every market at once, and the way you hear about it is email from professional services firms. Reviewing it and working out what applied to GIC took 10 to 20 per cent of the team’s time. The cost was not only the hours: on a volume that size, the risk is that the one change that mattered is the one nobody got to.
Made for the teamsthat cannot afford to miss a change
Tax teams
Rules that change in every market you hold a position in, arriving faster than one person can read them.
Compliance and legal
The same shape of work: a stream of advisories where missing one matters more than reading all of them.
Teams on an advisory retainer
Professional services firms send updates by email. Working out which ones apply to you is the job nobody scheduled.
The system reads the incoming advisories, categorises each one, is checked against the team’s own labels, and confirms nothing relevant was missed.
Take in the volume
Advisory email from professional services firms, arriving across every market the fund is exposed to.
Categorise each one
The model files each item, so the team starts from a sorted set rather than from an inbox.
Check it against people
Its categories are compared with the team’s own. Where the two disagree is what the accuracy is tuned on.
Confirm nothing was missed
A completeness check over the monitoring, because the failure that matters is a change nobody saw.
The results
The results, as the source reports them. It publishes no figures for them, so none are quoted here.
Less time on email
The review that took a tenth of the team’s week takes less of it.
Nothing relevant missed
Completeness is checked rather than assumed.
Tuned against judgement
Accuracy is measured against what the team itself decided, not against a benchmark.
What we can buildfor you
The same pattern, built around your desk. Each idea comes from a service we already run.
Know-your-customer and onboarding
Reads identity documents and company filings, cross-checks them, and prepares the file. An analyst reviews the exceptions instead of every case.
Transaction monitoring
Models trained on your own history rank alerts by how likely they are to matter, so investigators start with the cases that count.
Client and adviser questions
Answers drawn from your product terms and policies, with a clear handover to a licensed person when a question becomes advice.
Reconciliation and reporting
Matches statements, invoices and ledgers, and drafts the routine report. Your team owns the breaks and the final numbers.
Your team’s version of thisstarts with one inbox
Tell us the stream of information somebody is reading by hand, and we will tell you plainly whether AI is the right tool.