An office move used to mean a broker and weeks of email threads. In 2026 most of the planning can be drafted by AI tools a startup already pays for, such as ChatGPT, Claude or the assistants built into Google Sheets and Excel, which can turn a vague goal like “we need an office by spring” into a dated plan with a budget attached in one afternoon.
The same tools make confident mistakes with numbers and legal language, so each section below covers what to hand to AI and what a person should still check.
1. Work Backward From the Move Date
Start with the date the team needs to be working from the new space, then ask an assistant to build the schedule in reverse. A prompt as simple as “We need to move into a 2,500 sq ft office in Manhattan by March 1. Build a week-by-week plan working backward, including search, tours, negotiation, legal review and setup” will produce a usable first draft.
General-purpose models tend to underestimate how long a commercial lease takes, so give the assistant real benchmarks to build the plan around. Tandem’s first-half 2026 data for New York shows a median of 49 days from first tour to signed lease and one to three weeks from signing to move-in, so its planning rule for a move-in-ready suite is eight to ten weeks from first tour to move-in. The search itself starts earlier, because founders spend about 11 days browsing listings before that first tour, which puts the full time from starting the search at roughly ten to twelve weeks. Anything that needs construction should be scheduled months earlier.
Once the plan exists, paste it into a calendar tool or project board so the deadlines for decisions, such as approving a shortlist or signing a letter of intent, have owners.
2. Forecast Headcount and Space
Office size should follow the hiring plan for the next 12 to 24 months, not today’s team. If your hiring plan lives in a spreadsheet, Gemini in Google Sheets or Copilot in Excel can project headcount by quarter and convert it into square footage.
A common starting point is 150 square feet per person, which covers desks, meeting rooms and shared areas. Engineering-heavy teams can often work with slightly less, while client-facing teams usually need more for larger meeting rooms and a space to host visitors. Ask the assistant to show three scenarios (current plan, faster hiring and a hiring freeze) so the space decision does not rest on a single forecast.
In Tandem’s New York discovery calls, 94% of founders asked for room to double their headcount, so model the larger scenario before ruling it out.
3. Build the Budget With Real Market Numbers
A language model asked “how much does office space cost in Manhattan” will often return national averages or outdated figures, so give it current local data and let it handle the math.
For startup-scale offices, Tandem’s H1 2026 report on what New York startups pay per person is a practical source. It puts the median asking rent for Manhattan suites under 5,000 square feet at $64 per square foot per year and shows a U-shaped curve in all-in monthly cost per person: about $901 for teams of one to five, $825 for six to ten, $656 for 11 to 20 and $774 for 21 to 50. Paste figures like these into a sheet and ask the assistant to calculate total monthly cost for each headcount scenario from step 2.
Then ask it to add the costs founders tend to forget, such as the security deposit, furniture, internet installation, moving services and any buildout. The same report notes that a furnished, move-in-ready suite eliminates a $15,000 to $50,000 buildout conversation, so it is worth modeling a furnished option next to an unfurnished one.
4. Compare Listings and Proposals Side by Side
By the time a team has toured four or five spaces, the proposals rarely match in format. One quotes rent per square foot per year, another quotes a monthly total, and a third includes utilities while the others do not.
Upload the proposals or letters of intent to an assistant and ask it to extract the same fields from each into one table: rent, lease term, free rent, security deposit, escalations, what is included, tenant improvements and move-in date. Then ask it to calculate the effective monthly cost over the full term for each option. A prompt such as “Put these five proposals into one table with identical columns, convert all rent to monthly dollars and calculate the total cost of each lease over its full term” keeps the output consistent from one round of offers to the next.
Check the table against the source documents before sharing it with co-founders. Extraction errors are common with scanned PDFs and with numbers that appear in footnotes or side letters.
5. Do a First Lease Read With AI, Then Hand It to a Lawyer
AI tools are good at summarizing a 40-page lease into plain language and flagging clauses that deserve attention, including personal guarantees, assignment and subletting rights, renewal options, operating expense pass-throughs and restoration obligations at the end of the term. A first pass like this helps a founder arrive at the attorney meeting with specific questions, which shortens the review.
Accuracy is the risk here, as the 2023 case Mata v. Avianca showed when a federal judge sanctioned two New York lawyers who filed a brief citing court decisions that ChatGPT had invented. A lease summary can fail the same way by describing a clause that is not there or missing one that is. Treat the AI read as a briefing for your lawyer, and let the lawyer’s review decide what you sign.
6. Plan the Layout and the Move Itself
Assistants that accept images, including ChatGPT and Claude, can read a floor plan from a listing and give a rough count of how many desks and meeting rooms it could hold. That is enough to test whether a space fits the team before anyone pays a designer, as long as the final layout is checked against real measurements.
The move itself is a long list of small dependencies that an assistant can organize well. Ask it to produce a move-week checklist covering internet installation (which can take weeks to schedule in older buildings), building access cards, certificate of insurance requirements for movers, furniture delivery windows and IT equipment setup. Ask it to flag which items need to be booked first.
7. Know Where AI Stops Helping
Tours tell a team things no listing shows, such as street noise or how much light reaches the desks in the late afternoon. Negotiation depends on knowing which landlords will accept a shorter term or a smaller deposit, and that knowledge comes from relationships more than from data. The trade-off between cost and commute is also one only the team can weigh, and nothing should be signed until the lawyer has read the lease.
