If you’ve used Charli for policy research, you already know the basics: ask a question, get an answer, move on to the next client call. That’s been genuinely useful, but it’s also been limited in one important way. Until now, Charli has worked the way most AI tools work: she finds the closest matching answer to the question you’ve asked and gives it back to you. Fast, accurate, but fundamentally a lookup.

That’s changed. Charli has moved to agentic reasoning, and if you’re not familiar with the term, it’s worth fifteen minutes of your time to understand, because it changes what you can actually ask her to do.

Lookup vs. reasoning: the difference that matters

A traditional AI assistant treats every question in isolation. Ask “what’s the max LVR for this postcode with Lender X?” and you get a single, accurate data point. That’s helpful when your scenario has one variable. But broker scenarios almost never have one variable. They have five, six, sometimes ten, income type, employment structure, postcode, security type, existing debt, SMSF involvement, alt-doc requirements, all interacting with each other, and all relevant to which lender actually fits.

Agentic reasoning means Charli no longer treats your question as a single lookup. She works through it the way a senior credit assessor would: identifying each condition in your scenario, checking it against policy, then cross-referencing it against every other condition before landing on guidance. If a client’s self-employed income structure affects their serviceability calculation, and that serviceability calculation interacts with an SMSF security requirement, and that security requirement has postcode-specific LVR cap, Charli now works through that chain the way an experienced assessor sitting across the desk from a BDM would, rather than answering each piece in a vacuum.

Why this matters for the scenarios you’re actually working

Most of your day-to-day research isn’t simple. You’re not asking “what’s CBA’s LVR policy” in isolation, you’re asking it because you have a specific client with a specific, often messy, combination of circumstances. The value of an AI policy assistant isn’t in how fast it can retrieve a single fact. It’s in how well it can hold multiple facts in its head at once and tell you what they mean together.

This is the gap agentic reasoning closes. Where the old model could tell you what’s true, the new model can tell you what’s true for this client, in this scenario, given everything else you’ve told it. That’s a meaningfully different kind of output, and it’s the difference between an assistant that answers questions and one that actually helps you assess deals.

A practical example

Take a scenario most brokers will recognise: a client who’s self-employed, drawing income through a trust structure, wanting to use an SMSF to secure an investment property, in a postcode that several lenders treat as “regional” for LVR purposes. Under the old model, you’d need to ask Charli four or five separate questions, one on alt-doc income policy, one on SMSF lending criteria, one on postcode LVR caps, one on trust structure requirements, and then do the cross-referencing yourself. Useful, but still manual stitching.

With agentic reasoning, you describe the full scenario once. Charli identifies each policy-relevant factor, checks it independently, and then reasons across all of them together, flagging, for example, that while Lender A’s SMSF policy looks attractive on paper, their postcode classification for this specific property knocks their effective LVR down below what the client needs, while Lender B’s slightly more conservative SMSF stance is actually the better fit once the postcode and income structure are both accounted for. That’s not a lookup. That’s an assessment.

What this means for your day-to-day workflow

In practice, this shift changes how you should be using Charli. Instead of breaking a complex scenario into a series of narrow questions, you can now describe the whole picture, client circumstances, security, structure, and goal, in one go, and let Charli do the cross-referencing that would otherwise eat into your prep time before a BDM call or client meeting.

It also means the guidance you get back is more directly usable. Rather than a list of individual policy facts you still need to weigh against each other, you get reasoned, scenario-specific guidance,  the kind of output that’s ready to inform a client conversation, not just a research note.

How this changes the questions worth asking

Once you understand that Charli is reasoning rather than retrieving, it changes what’s worth bringing to her in the first place. A narrow, single-fact question is still answered well, but it’s no longer the best way to use the tool. The better instinct is to bring her the kind of question you’d genuinely take to a senior colleague: “Here’s everything about this client and this deal, where does it land, and what would I need to watch?” That’s a question a lookup tool can’t meaningfully answer, because there’s no single fact being requested. It’s exactly the kind of question agentic reasoning is designed for.

This also changes what a “good” research session looks like. Previously, a thorough broker might run four or five queries on a complex file and feel confident they’d covered the bases. Now, one well-described scenario can surface the same ground, plus the interactions between policy clauses that a series of narrow queries might never have prompted you to check in the first place, simply because you didn’t know to ask.

A note on what hasn’t changed

It’s worth being clear about what agentic reasoning doesn’t replace. Charli isn’t making credit decisions, and she isn’t a substitute for your own judgment, your compliance obligations, or your relationship with a lender’s BDM team on genuinely unusual files. What’s changed is the quality and completeness of the research that sits underneath your judgment,  giving you a more reliable starting point, not a final answer that removes the need for broker expertise. The best results still come from a broker who knows how to read the output critically and knows when a scenario genuinely warrants a human conversation alongside the AI guidance.

The bigger picture

Brokers have spent years being told AI will “save time.” Time-saving is real, but it’s only half the story. The other half is decision quality, and that’s where reasoning over lookup makes the biggest difference. A faster wrong answer isn’t progress. A faster, more complete, properly cross-referenced answer is.

That’s the shift behind Charli’s upgrade. Over the next several posts in this series, we’ll walk through exactly how this plays out across real scenario types, alt-doc, SMSF, multi-lender comparisons, postcode-sensitive deals, and what it means for the time you reclaim and the submissions you put forward with more confidence.

If you haven’t tried describing a full, multi-factor scenario to Charli yet rather than breaking it into separate questions, that’s the place to start. Try it on your next complex scenario at cynario.ai/charli.

Author – Charli