One of the many emails that will never be opened
Your inbox holds dozens of them again today: unread emails from senders you don't recognize, each trying to sell you something. A single glance at the subject line reveals that the sender knows nothing about you or your company. The person who sent that forever-unopened email will log it in their CRM not as a lost deal, but as a "no response." Yet you have already decided not to buy, and the decision took less than a second.
In a 2024 Gartner survey of B2B buyers, 73% of respondents said they actively avoid companies that send irrelevant sales messages. Not "ignore," but "avoid." A careless message that shows no grasp of the customer does not end with a single unanswered email. It can disqualify a supplier from deals that haven't even begun. Gartner's report puts it more bluntly: "Bad prospecting actively damages relationships with potential customers."1
The game is decided before the first conversation
There is another uncomfortable finding. Bain, surveying enterprise-software buyers alongside LinkedIn in 2026, reached a conclusion that compresses into a single sentence: "Most seemingly 'lost deals' were never winnable because the vendor was not on the buyer's initial shortlist to begin with." Bain calls that shortlist the "Day 1 list"2 — the companies already in the buyer's head the moment the purchasing process begins. BCG's findings point in the same direction. In its four-country survey of small and medium-sized enterprise (SME) buyers (2023), 76% of purchases stayed within the network of vendors the company already did business with. Even for a first-time product purchase, more than half of the buyers chose a company where they already had a point of contact.3
6sense reached the same conclusion — though this should be weighed knowing the company sells marketing software for spotting in-market accounts early. In its global buyer study (2025), the winning vendor was already on the Day One list in 95% of deals.4
The message of these surveys converges on a single point: the buying decision is made before you even step onto the field. The list is drawn up earlier than you think, and it is built on existing relationships. By the time the seller hears about the deal, the buyer has all but made up their mind.
Most people stop here, shrug, and conclude: "So B2B is a game played among people who already know each other." But that does not mean the Day 1 list is forever out of reach. The same 6sense report carries this sentence: "Our prior research showed that B2B buyers select a favored vendor before engaging sellers — and that pre-contact favorite wins the deal roughly 80% of the time."4 Read it from the other side: the favored vendor is chosen before first contact. The goal then becomes clear. Armed with proper research, approach potential customers early, get noticed, keep the conversation going, and start with a small deal. That way, long before the next buying process begins, your company's name is already on that Day One list.
What "proper research" actually means: three questions
Over the past decade and more, the team at A Piece of Map has discovered new B2B partners, advanced discussions, and pushed them through to signed contracts time and again. Looking back, the pre-contact research that actually delivered results shared a consistent shape. When we walked in with answers to three specific questions, we could steer the conversation with confidence. When we skipped them to save time, we rarely saw results.
1. The market — what is happening in it right now? Before studying any specific company, you must first understand what the market you are entering actually looks like. Who operates in it, how it divides into segments, and what has changed over the past year. Who acquired whom, who partnered with whom, and what regulatory or technological shifts are moving the market's money. These are the questions an in-house team or a commissioned market study typically answers. From a sales and business development perspective, you read that information to determine whether there is a place for you in this market at all, which sub-segment your product belongs in, and whom the market's buyers will compare you against. Skip this research, and every subsequent step may head down the wrong track.
2. Target companies — where to focus first? Once the market is mapped out, you identify the key companies in each target industry within it. Then you assess each company's business situation: what it sells, who its customers are, what it announced recently, and what problem it is trying to solve this year. In truth, half of answering these questions is an exercise in subtraction. Cross off your competitors, and cross off the companies so firmly tied to another vendor that no opening remains. Because shortlists tend to form out of existing relationships, spending your early energy on an account where an incumbent already holds a solid foothold is rarely wise. A more workable path is to build references in the new market first, then position yourself as a point of comparison against that incumbent.
3. The entry path — which partner can introduce you to the target (the warm path)? The remaining homework is figuring out how to actually approach the companies that survived the subtraction and made the target list. Yet in new-market research, almost nobody considers this question. Recalling that more than three-quarters of purchases happen within existing trading relationships, you realize how hard it is to knock on a door empty-handed. Make the question concrete: where does my world already intersect with theirs? In other words, find the path. Look for a mutual partner or a common supplier where your existing relationships could naturally produce a warm introduction. It is fine if no names come to mind immediately. Useful clues lie scattered across published press releases, regulatory filings, and partner pages. Assemble those materials, and the odds are good the answer is in there.
But who is going to do this work?
By the US Census Bureau's own count, 78% of all American businesses have no employees at all.5 They are one-person operations that must handle everything themselves. Things are not much different at slightly larger companies, or at workplaces of thirty to fifty people. In a small company, "the sales team" is often two or three people juggling market research, marketing, contracts, delivery, customer support, and much else. Under the title of business development manager or sales manager, the work of several departments is compressed into a single role.
Salesforce's 2024 study puts numbers on that workload. Sales reps spend only about 30% of their workweek actually selling. The other 70% is tied up in desk work: writing proposals, getting internal approvals, preparation and planning, and administrative tasks. The report files this under "Productivity Lags as Nonselling Tasks Dominate Time" and focuses on finishing that work faster to lift productivity.6
But what if we turned that thinking around? Instead of forcing a salesperson to finish non-selling work faster, what if we handed the work itself to an AI agent as a division of labor? Under the report's own task classification, answering the three questions above falls under prospect research, preparation and planning, and lead/opportunity prioritization — about 26% of a rep's working time. Divide that work with an AI agent, and the salesperson converts substantial hours back into actual selling. If an organization is so short-handed that it has never managed to dedicate hours to pre-contact research, this shift in thinking — a division of labor with AI — may surface opportunities with customers it never previously considered.
Pulling an all-nighter with an AI agent
When a one-person or small company decides to offload its market research to AI, the odds are good they will find themselves wrestling with a general-purpose AI agent all night. There are two reasons for this.
First, babysitting the AI is an added cost. The Work AI Index (2025–2026), a survey of six thousand digital workers in three countries by Stanford, Berkeley, and Glean, measured a phenomenon it calls "botsitting"7 — "feeding it context, supervising its output, debugging its mistakes, and cleaning up after it." The finding: workers spend 6.4 hours a week on this — most of a full working day, every week. More than a third of AI sessions "fail outright, requiring a full restart or substantial rework," and the report also found that "for every hour a worker spends getting useful output from AI, they spend roughly another hour making it usable."
Second, AI can make up answers at the moments that matter. A team of medical researchers had GPT-4o write six literature reviews on mental health topics. They then checked all 176 citations in the reviews, one by one, against academic databases (JMIR Mental Health, 2025).8 The result: 35 of them (19.9%) were fabrications citing papers that do not exist. The researchers had deliberately chosen three topics of differing public familiarity. They hypothesized that the model would fabricate more on topics with thinner public records, and the results ran exactly as predicted. On the best-known topic, the fabrication rate stayed at 6%; on the most obscure ones, it climbed to 28–29%.
Come back to the pre-contact research questions we established earlier. Take the second question: researching target companies. Not every B2B company is researching Apple or Samsung, where information overflows online. Most of the time, the targets are mid-sized and smaller companies in a market you are entering for the first time, in a field you have barely heard of. The study above shows that this is precisely where the generic AI tool is weak.
The third question — mapping relationships between companies — poses an even harder challenge for AI. It is not simple retrieval, but assembly. Finding out who can introduce you to whom is not information written on a single website or in any single article. It is an answer that only surfaces when announcements, partner pages, and filings from different sites are laid side by side and connected. Fitting those pieces together through the night is punishing work.
A tool that makes up for what AI lacks
Running into these difficulties while working with an AI agent is no reason to sink into skepticism and abandon the technology. Babysitting, fabrication, and failed assembly share one root cause: they happen when everything from setting the research direction to executing the search is entrusted to a single machine without a consistent plan or guide. What about putting another AI by its side — one that sets the direction and handles the verifying and evaluating?
APiM Galaxy (galaxy.apieceofmap.com) was built to bypass these difficulties and make AI carry out business research the way business development and sales actually need it done. It connects via MCP (Model Context Protocol) to the AI you already use (Claude, ChatGPT, Claude Code) and runs the research for a B2B company entering a new market. That means answering the essential questions set above: the market environment, the target companies, and the entry path. Because the research direction is already built in, a single prompt is enough for it to conduct the search, write the report on its own over ten or so minutes, and deliver the results. As it answers each question, it has the MCP check each cited source against the page it names, and flag what it can't confirm rather than hide it. The work a person used to piece together overnight? The AI draws the map from company-to-company relationships checked against their sources, with unconfirmed ties flagged as such.
Adopt APiM Galaxy, and a company's business development and sales teams win back the time to leave their desks, meet, and talk with more customers in person. Beyond that, it genuinely cuts the cost and waiting times of reports bought from professional consulting firms. A report that costs roughly $15,000 to $60,000 per engagement and takes six weeks to nearly two months to arrive can be replaced in ten or so minutes, at practically no cost, simply by installing one MCP on the AI agent you use every day.9
What people should concentrate on
Where should people spend the hours the AI tool wins back from research? Two experiments offer a hint.
The first experiment (2019), conducted in fintech phone sales, highlights a selling point only a human can reach.10 When customers received a call from an AI chatbot that did not reveal its identity, it performed just as well as proficient salespeople. It even sold four times as much as inexperienced human reps. But when researchers told customers in advance that the caller was a machine, purchases collapsed by 79.7%. It happened "despite the objective competence of AI chatbots." Same machine, same performance. Yet the moment customers knew it was a bot, they judged it less capable and less empathetic. It is less a question of ability than of genuine customer psychology: people simply do not open up to machines. Hence the conclusion: in the seat where trust and relationships are built, a person must still stand.
The second experiment (2024), at a B2B aluminum retailer, pins down more concretely the kind of work humans do better.11 When the pricing of routine, repeated quotes was handed to an algorithm, profits rose 11% compared to when salespeople set them. But on special quotes with unusual or complex characteristics, the salespeople beat the algorithm. The best-performing arrangement in the experiment handed everything to neither side. It used a division of labor that sent 68% of quotes to the algorithm and the out-of-the-ordinary 32% to people, earning 7.8% more profit than the salespeople working alone.
The same preference shows up on the buyer's side. In the global B2B buyer survey McKinsey published in 2026, buyers preferred digital self-service when transactions were frequent and repetitive. But on large, infrequent orders — even reorders — they strongly preferred to talk with a person.12 The biggest reason buyers leave a supplier is likewise the relationship, not the product: when every point of contact tells them something different, or when they cannot reach someone who actually knows them.
Lay these results over our earlier story, and the line dividing the labor comes clearly into focus. Work that repeats and can be verified against sources, the machine does well. In work that is out of the ordinary, where judgment hangs on the call, and where the other side's trust must be won, the human wins. The three questions we set earlier — what is happening in the market, which companies to focus on, and who can make the introduction — all sit on the former side, the side that can be handed to the machine. But the step of walking into the meeting that introduction opened, the judgment required to negotiate the terms of an unusual deal, and the hours spent bringing the other side to decide "let's work with this company" — those were never the machine's job to begin with.
Go back to the inbox where we started. The unopened email was the one sent without research. The game was already decided on the Day 1 list before the first conversation even began. There is no shortcut onto that list. Read the market, choose the targets, find the introduction paths. Do the pre-contact research, and build relationships early on top of it. Only one thing has changed: the repeatable, verifiable part of that research can now be shared with AI. With the hours won back, people should concentrate on what only people can do: meeting more potential customers in person, starting with small deals, and building trust. That way, on the day the next purchase begins, your company's name is already on that Day One list.
APiM Galaxy
One question in. A full market-entry strategy report out.
Connect the AI you already use — Claude, ChatGPT, Claude Code — and share the research that decides the Day One list: the market, the targets, and the way in.
Explore APiM Galaxy →Footnotes
- Gartner press release, June 25, 2025. Survey of 632 B2B buyers, August–September 2024. https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-sales-survey-finds-61-percent-of-b2b-buyers-prefer-a-rep-free-buying-experience ↩
- Bain & Company with LinkedIn, "Raising the Odds on a Deal," June 19, 2026. Fieldwork by NewtonX, February 2026; 750 buyers responded. The survey covers two enterprise-software sectors (human capital management / cloud data platforms), and Bain states that generalization to other industries is still being tested. The same page also says "around 90% of buyers purchase from their Day 1 list," but no sample or method is given, so this piece does not quote it. https://www.bain.com/insights/raising-the-odds-on-a-deal-how-likelihood-to-buy-rewires-b2b-growth/ ↩
- BCG, "B2B Customer Dynamics: The Weather Ahead," September 28, 2023. Survey of 2,700+ SMB purchase decision-makers and influencers across France, Germany, the UK, and the US, fielded August 2023. The survey question reads, "Did your company have an existing relationship with the vendor you ultimately purchased from?" — "relationship" here means an actual trading relationship, not mere name recognition. https://www.bcg.com/publications/2023/b2b-customer-experience-trends-in-uncertain-future ↩
- 6sense, "2025 B2B Buyer Experience Report." Global study of nearly 4,000 B2B buyers. https://6sense.com/science-of-b2b/buyer-experience-report-2025/ ↩
- US Census Bureau, "The Steady Rise of the Nonemployer Business," July 30, 2025. https://www.census.gov/library/stories/2025/07/nonemployer-business-growth.html ↩
- Salesforce, State of Sales, 6th edition, p. 13. Survey of 5,500 sales professionals across 27 countries, fielded March–April 2024. The 30% selling time comes from the report's eleven-category workweek chart; the 26% in the body is the sum of researching prospects (9%) + preparation and planning (9%) + prioritizing leads and opportunities (8%). https://www.salesforce.com/resources/research-reports/state-of-sales/ ↩
- Work AI Institute (Glean), "The Work AI Index: Global." Survey of 6,000 full-time digital workers in the US, UK, and Australia, December 2025–January 2026, nationally representative sampling, self-reported. The study covers knowledge work in general — the figures are not specific to sales roles. https://www.glean.com/work-ai-institute/reports/work-ai-index ↩
- Linardon, J., Jarman, H. K., McClure, Z., Anderson, C., Liu, C., & Messer, M. (2025). Influence of topic familiarity and prompt specificity on citation fabrication in mental health research using large language models: Experimental study. JMIR Mental Health, 12, e80371. https://doi.org/10.2196/80371 (PMID 41223407 · PMC12658395). ↩
- Based on actual transaction prices published in US federal procurement records (FPDS · USAspending). ↩
- Luo, X., Tong, S., Fang, Z., & Qu, Z. (2019). "Frontiers: Machines vs. Humans: The Impact of Artificial Intelligence Chatbot Disclosure on Customer Purchases." Marketing Science, 38(6), 937–947. Randomized field experiment with 6,255 customers. https://pubsonline.informs.org/doi/10.1287/mksc.2019.1192 ↩
- Karlinsky-Shichor, Y., & Netzer, O. (2024). Marketing Science, 43(1), 138–157 (DOI 10.1287/mksc.2023.1449). https://pubsonline.informs.org/doi/10.1287/mksc.2023.1449 ↩
- McKinsey, "The surprising economics of B2B growth," June 2026 — the 2026 Global B2B Pulse. Fielded December 2025; 3,664 respondents across 13 countries. https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/the-surprising-economics-of-b2b-growth-the-new-survival-threshold-and-what-it-takes-to-thrive ↩
