Using AI in a B2B business, one question at a time.
Real questions from B2B business owners and managers, collected in October 2026. We've kept the wording close to how they were asked and answered each one the way we would across a desk – what we think, why, where it goes wrong, and what we'd do first.
The opinions are ours and marked as such. Nothing here is legal, financial or security advice for your particular situation.
Strategy and who leads it
Who owns AI, how it starts in a business where everyone's already busy, and what to protect while you change things.
An AI strategy for a company where nobody owns it yetWhat does an AI strategy look like for a company just starting out, where nobody owns it and the department heads are fully busy in their own roles?
Department-led, IT-led or management-led? Split it three waysShould AI be led by each department, centralised through IT, or kept with management (who can get bogged down)?
The best AI lead is rarely the most technical personWho makes the best AI lead or champion: the person with the most technical knowledge, or someone with leadership ability and vision but no credentials?
Designing one automated process from several people's expertiseCan you bring different people's expertise together under one head to design an automated process?
Tailored agents for specific problems, or one general tool?For smaller businesses, is it better to start with tailored agents for specific problems, or a general generative or agentic tool that covers lots of things?
Redesigning the business from the top down without losing what makes it yoursHow do you redesign the business from the top down safely, and what guardrails protect the company DNA while you do it?
From experiments to something people actually use
Prototypes are cheap now. A finished tool a team relies on, with an owner and one version, is the hard part.
Prototyping is easy now. Finishing one thing is the hard partPrototyping is easy now, so how do you balance encouraging new ideas against taking one through to a fully implemented, maintained solution, rather than always building and testing?
Who gets to build live tools, and who builds concepts onlyHow many people are allowed to build live tools versus just concepts, given compliance and data integrity, so not everyone is shipping agents randomly?
Every agent needs an owner. What that owner actually doesDoes each agent have an owner, or is there a team looking after different areas?
One source of truth, so an update reaches everyone using an agentHow do you set up agents so there is one source of truth and updates roll out to everyone using them?
Keeping agents on one platform, so staff aren't confused by silosHow do you keep agents on one platform so you do not end up with silos that confuse staff?
Now everyone has an AI assistant, how do you keep salespeople selling?Now Claude is rolled out to the whole team, how do you keep salespeople selling rather than building the next cool thing?
Stopping low-value tasks drifting up to senior peopleHow do you stop lower value tasks drifting up to senior people without anyone noticing?
Busier than ever because you can do so much yourself: when to delegate and when to use AIAs a business owner I am busier than ever because I can do so much more myself, so when should I delegate and when should I use AI?
Data security, privacy and access
What AI should be allowed to see, how to restrict it, and how to stop staff and vendors putting the company at risk.
What data AI should be allowed to see, and what it shouldn'tWhat data should AI have access to without risking sensitive data, given it is more useful the more it can see?
Protecting IP, company data and the financialsHow do you protect IP and the security of company data and financials?
How much customer information our agents can see, how it's restricted, and how sure we areHow much sensitive customer information do your agents have access to, how did you restrict it, and how confident are you that it is restricted?
The policies and controls you need when access is open at every levelWhat policies and controls do we need, especially if access is open at all levels? How reliable are agents on privacy?
Stopping staff putting company data into AI tools on their own initiativeHow do you stop staff using AI on their own initiative and putting company data at risk?
Leaders who won't give AI "too much", and whether an enterprise licence already protects youHow do you deal with leaders who do not want to give AI too much information, and do enterprise licences like ChatGPT Enterprise already protect the business? Also, IT's concerns about AI accessing company folders.
Your systems on someone else's infrastructure. What if the vendor shuts down?How do you manage the risk of your systems sitting on another company's infrastructure, and what happens if that vendor shuts down suddenly?
Data quality, knowledge and context
AI is only as good as what it can see. What to capture, how to get knowledge out of people's heads, and what RAG and MCP actually mean.
Working out what data to capture, in sales for example, and keeping it cleanHow do you work out what data you need to capture (in the sales process, for example) for AI to be useful, and how do you keep it clean?
Getting the knowledge out of people's heads so everyone can use itWhere a lot of knowledge sits in people's heads, how do you get it out so everyone can use it?
Your internal context layer, and what RAG and MCP actually areHow do you manage your internal context layer, through MCP or some native context control? And related, what are RAG and MCP?
When AI answers are frustrating, is it usually a context problem?When AI answers are frustrating, is that usually a context problem?
A simple way to validate what AI produces, and keeping the guardrails up as models changeIs there a simple way to validate what AI produces, and will internal guardrails keep up as models change? How do you bridge the gap between technical human requirements and what the AI puts out?
Training, adoption and change
Getting people trained, out of the day to day for long enough to learn, and past the negativity.
Training for all staff: what we'd recommend and how to roll it outWhat training programs, online or structured, do you recommend for all staff, how did you roll training out, and what is the agent training you mentioned your team doing?
Getting people out of the day to day, dealing with negativity, and helping them accept the changeHow do you get people out of the day to day to spend time on AI, deal with negativity, and help them accept the change?
Cost, budget and results
What to budget, whether the bill climbs once agents are embedded in the work, and how to measure the return without kidding yourself.
How to budget for AI, and whether costs climb once agents are embeddedHow do you budget for AI (what percentage of profit), and do costs suddenly climb once agents are embedded in workflows? How much budget does a change like this need?
Measuring the return on AI: the numbers worth trackingHow do you measure ROI, and how has AI changed One NZ at a high level on profitability, customer satisfaction, headcount and staff satisfaction?
Tools and specific examples
Choosing between the assistants, what an agent actually needs, build or buy, contracts, and customers who'd rather talk to a person.
Choosing between ChatGPT, Claude and the rest, and switching without losing what they've learnedHow do you choose between ChatGPT, Claude and the rest, and how do you manage switching between tools given what each one learns about you?
What you need to set up an agent, and what you don'tWhat tools do you use to set up agents?
The agents worth building first, and how our favourite worksWhat is your favourite agent and how does it work? Which kinds of agents should a business build first?
What a weekly AI show-and-tell looks likeWhat kinds of things does your team present at the Friday meetings?
Build or buy, with a contract management system as the exampleFor the contract management system, how did you decide to build rather than buy, and did you consider ERP integration?
Getting AI to flag problem clauses in contracts, and how far to trust itHow do you get AI to assess risk, for example flagging problem clauses in legal contracts?
Getting customers past the idea of talking to an AI rather than a personHow did you get customers past the negative view of talking to an AI rather than a person?
Workforce and pricing
Hiring when AI does a lot of the entry-level work, and whether being efficient should turn into lower prices.
Hiring now that AI does a lot of the entry-level workHow are you handling new hires now that AI can do a lot of entry-level work?
If AI makes you cheaper to run, do you cut prices to win share?Does being the most AI-led provider mainly mean passing efficiencies on as lower prices to win market share?
Where the answers come from
We run AI inside B2B sales teams for a living, mostly in industrial and professional firms with a handful to a few hundred staff, so the answers lean on what we've seen work and fail in real businesses. Where something is our opinion we say so. Where a question needs a lawyer, an accountant or your own security people, we say that too. We don't name a tool as the answer, because the tool is rarely the problem.
If a question of yours is missing, send it to us and we'll add the answer.

It starts with a discovery meeting.
A conversation about your market, what's in your CRM today and what a good customer looks like for you. If there's a fit, you get a written proposal.