PacificWest 2026

PacificWest 2026 is the joint conference of Greater Vancouver REALTORS® and the Fraser Valley Real Estate Board, and the largest gathering of REALTORS® in British Columbia. Two days at the Vancouver Convention Centre East, September 28 and 29, with six self-directed PDP hours for GVR and FVREB members who attend in full.

Local Logic is there for both days. On September 28 our VP Marketing, Max Leblond, takes the Mini Stage to talk about what happened to local expertise in the year every client started arriving with an AI answer already in hand.

>> Book a meeting with us in Vancouver

Why Local Logic is at PacificWest 2026

Something changed for BC REALTORS® this year. Clients now open the conversation holding output from a tool that has never been to the street. A seller pushes back on price mid-presentation because two AI tools handed them numbers. A buyer describes a neighbourhood before ever standing on it. Both of them sound informed, and both of them are often wrong.

Statistics Canada puts generative AI use among workers in finance, insurance and real estate at 59.2%, second only to professional, scientific and technical services. Consumer trust moved the other way over the same stretch: the share of buyers who trust AI to help them find a home fell from 30% to 16% in a single year, according to Cotality’s survey of buyers in the US, Canada, the UK and Australia. Almost everyone in the industry is using it. Almost nobody on the other side of the table believes it.

That gap is the opportunity. Knowing the area used to be the differentiator, and a machine now hands a convincing imitation of that knowledge to everyone, instantly and free. What stays scarce is being right, and being able to show why. Local Logic quantifies everything outside the four walls: walkability, transit access, amenity density, noise, park access, school proximity, demographics and climate risk, measured for any address in Canada and the US.

PacificWest draws the people this lands on hardest: REALTORS® from Greater Vancouver and the Fraser Valley, managing brokers and brokerage owners, team leads and mentors, board and MLS technology teams, and the marketing leads deciding what neighbourhood data belongs on their listing and community pages.

Our session: when everyone has the same AI, being right is the only edge left

Mini Stage, Monday September 28, 1:15 p.m. to 1:45 p.m. PT. Presented by Max Leblond, VP Marketing at Local Logic.

The idea that explains the gap is grounding: the difference between a model answering from memory and one that looks up the real number first. That difference is invisible in the answer, which is exactly the problem.

We measured it. We put 490 neighbourhood questions covering 46 neighbourhoods to 7 AI models, asking each question twice: once from the model’s memory, once grounded in Local Logic data. The ungrounded answers produced 1,861 claims that contradicted the real data. Grounding cut contradicted claims by as much as five times on the cheaper models and roughly doubled the number of verified facts per answer. The full method and results are here.

The finding that should worry a working agent most: ungrounded accuracy fell as neighbourhoods got less famous, while grounded accuracy held steady across every tier. Ask a model about Yaletown and it performs passably. Ask about a street in Renfrew or South Marine and it starts filling gaps with invention. Independent researchers found the same shape. LocalBench, a benchmark of 14,782 questions across 526 US counties, reports that the best of 13 models managed 56.8% accuracy on narrative local questions and under 15.5% on anything numerical.

You will leave with a way to tell a grounded answer from a merely fluent one, the questions worth asking any AI tool before its output goes in front of a client, and a straight account of what happens to a listing presentation when the client’s numbers are wrong and yours are checkable.

See Local Logic’s data in action: Vancouver

The best way to understand grounding is to point it at the room the session is in. Every number below comes straight from the Local Logic platform for 999 Canada Place, the address of the Vancouver Convention Centre East, and for Dunbar about seven kilometres to the southwest. Nothing here is recalled or estimated.

Two addresses in one city, with near-inverted profiles. Ask a model to compare them and it will tell you downtown is busier, which is true and useless. The scores tell you by how much, on which dimension, and for which client.

Schools are where models guess most confidently and miss most often. The nearest secondary school to the Convention Centre is Alexander Academy on Dunsmuir Street, private, grades 10 to 12, about 560 metres away. The nearest public secondary is King George Secondary on Barclay Street, grades 8 to 12, about 1.8 kilometres. The nearest elementary is the New School at Coal Harbour on Broughton Street, public, kindergarten to grade 7, about 1.1 kilometres. Three different correct answers depending on what the client actually asked.

Climate is where the gap is widest. The Convention Centre address carries a precipitation rating of 96 out of 100, with annual precipitation modelled to move from 1,541 mm to 1,605 mm by 2050. Heat is the number that surprises people: 33 days a year above 27.1°C by 2050, against a heat rating of 44. Flood and wildfire ratings both sit at 1. A model working from the general reputation of Vancouver weather gets the rain directionally right and the heat completely wrong.

Nothing above required us to know Vancouver. It required a system that looks the numbers up before answering.