A field guide to running partner programs with technology companies, hyperscalers and marketplaces, GSIs, and resellers. The five steps stay the same. What you do at each one depends on the partner, and on how AI agents are changing partnering in 2026 and 2027.
Most technology is now bought through, with, or alongside a partner. Getting the program right is no longer a side project.
of global IT spend in 2025 goes to, through, or with channel partners.
Canalys, 2025 1of enterprise software spend already goes through cloud marketplaces, according to a survey of 833 buyers.
Futurum, 2026 8forecast yearly spend through hyperscaler marketplaces by 2030, up from $16B in 2023, with 59% expected to flow through channel partners.
Canalys / Omdia 2, 3less likely to churn: customers who use a product's integrations.
Crossbeam vendor survey, 2023 4About the numbers. A lot of partner research is paid for by the clouds and software vendors who benefit from it. Wherever that's the case, I've said so. Treat those figures as a sign of direction, not proof, and lean on your own data where you have it.
Every partner type gets paid in something different. Follow each line to its last stop: that's what the partner is paid in. In my experience, a program works when your offer helps them earn more of it, and stalls when it doesn't. If you only remember one idea from this page, make it this one.
Paid in: Retention
An integration that makes their product harder to leave, plus introductions to your customers that fill their pipeline.
Paid in: Consumption
Deals that draw down customers' committed cloud spend and count toward their sellers' numbers.
Paid in: Services
Billable work attached to your product, and an edge in large transformation deals.
Paid in: Profit
Margin on the deals they find, and increasingly recurring revenue from managing the customer afterwards. Both need protecting from your direct team.
These are the plays I'd run. Pick a partner type to see its playbook: each step has one move to make, one trap to avoid, one number to watch, what's changing for 2026–27, and a real company that did it.
Fewer, better partners. Every partnership costs focus, so pick the ones that share your customers.
Tick what's true. Use it before you recommend a partner to a customer.
0 of 6. Tick the traits this partner has.
Create joint value you can describe in one sentence, before anyone signs anything.
Two companies that haven't worked together will hit slow approvals, mismatched goals, and cautious teams. Plan for it:
Good first projects: a joint sales training session, co-written content, or a lightweight connector. What you learn about how they work is worth as much as the result.
Make it easy for their sellers to win without you in the room. Show, don't tell: a live demo on the customer's own data beats any one-pager.
Turn the partnership into named deals in named accounts.
Track what each side is actually paid on, then put more into what works.
Opinions from more than a decade of building partner programs. You may see it differently, and that's exactly what the call button is for.
GSIs
Global agreements are easy to sign and easy to ignore. The partnerships I've seen work had one thing in common: a practice leader inside the integrator whose targets moved when our joint business grew. Find that person before you sign anything, build the business case around their goals rather than yours, and fund their first few engagements if you need to. Logos on a slide don't move pipeline. People with targets do.
Hyperscalers
Most ISVs walk into a cloud provider with a product deck. The account teams I worked with cared about one thing: whether a deal drew down a customer's committed spend and counted toward their goals. Lead with that, in their language. Make the marketplace private offer the default way you transact, show up in their account planning, and be the partner who makes their quarter easier.
Technology partners
In AI especially, there's pressure to announce partnerships fast. I'd rather launch something small that people actually switch on than a big announcement nobody uses. Pick the one workflow where both products touch, ship it, and watch activation. Co-marketing works far better when you can point to real users, and it gives both sides a reason to invest in what comes next.
Channel
When you build a channel program from scratch, the temptation is to recruit widely and see who sticks. I'd start narrow: a handful of partners you work alongside closely. They'll show you how your product really gets sold, where deals stall, and what enablement they need. Write the playbook from what they teach you, then scale. Recruiting before you have a playbook just multiplies the confusion.
Every partner type
Most partner programs don't fail with the partner. They fail inside your own company, when direct reps see partners as competition for the same deal. Fix the incentives first: make sure your sellers win when a partner is involved, get a senior sponsor to back the program in public, and share partner wins internally as loudly as external ones. A partner team can't out-run its own sales compensation plan.
Most partnerships fade quietly before they fail. This is the quick check I use: rate one partner on three areas to see whether it's healthy, worth watching, or at risk, and what to do next. The scoring is a rule of thumb, not a formula. Run it every quarter.
Rate all three areas
Your result and a recommended next step will appear here.
AI agents are changing what partners sell, how they price it, and how customers find it. These are the five shifts I'd plan for, whichever partner type you work with.
Partners are adding AI and agents to every service line. Routine work such as migration and support gets cheaper as agents take more of it on, while new work opens up in designing, building, and running agents for customers.
82% of AWS partners already deliver some form of AI to customers. Omdia 2025, via AWS 5
Customers want partners to share the risk. Build offers where you and your partner are paid against a business result, and pitch them to line-of-business leaders, not only IT.
Outcome-based pricing doubled from 18% to 36% of engagements, while fixed-price work fell from 49% to 29%. 75% of line-of-business buyers say partner relationships are essential. IDC, via AWS 5
Cloud marketplaces now sell bundles that combine several vendors' products with professional services, billed by milestone or outcome. Buyers increasingly find them by describing their problem to an AI assistant. The winning offer is often assembled by three or four partners, and its listing has to answer use-case questions directly, for people and for AI.
AWS Marketplace added multi-product solutions, variable payments for services, and an AI "agent mode" for discovery in late 2025. AWS 5
Google Cloud's agent finder matches buyers to partner agents in plain language, and agents can be priced by outcome, such as per support ticket resolved. Google Cloud, Oct 2025 12
The strongest partners package their know-how into reusable, industry-specific agents and accelerators that they own and list on marketplaces. Help them build that IP on your platform: it raises their margin and makes your product harder to replace.
When customers pay for what they use, such as API calls, tokens, or compute, a signed contract only buys access. Revenue arrives as usage grows. AI is arriving faster than most customers can absorb it, so the partners who stay after go-live, help customers adopt, and find the next use case become hard to replace.
Helping customers adopt and manage Google Cloud accounts for 35% of the partner services opportunity around it. AI projects typically move from pilot to production in year three. Canalys, commissioned by Google Cloud, 2025 9
You'll ask partners to sell AI. Your own team should be using it first. Here's why it's worth doing, when it makes sense, what it takes, and a path from first experiment to an agent-run partner operation.
Partner work is heavy on coordination: researching accounts, matching overlaps, chasing referrals, updating portals, answering the same product questions, and building quarterly reviews. That's time not spent with partners.
Sellers spend only about 40% of their time selling. Those using agents expect research time to drop 34% and email drafting 36%. Salesforce State of Sales, 2026 6
There's a credibility reason too. Partner leaders who run their own team on agents can tell partners what works from experience.
A task is a good candidate when all four are true:
Keep people on relationship moments, negotiations, and strategy.
Gartner expects over 40% of agentic AI projects to be scrapped by 2027 because of cost, unclear value, or weak risk controls. Start where the value is obvious. Gartner, June 2025 7
Set guardrails from day one: a person approves anything sent to a partner or customer, agents only see the data they need, and every action is logged.
Research prospective partners and analyse account overlap reports into a ranked shortlist.
Draft joint solution briefs, marketplace listings, and use-case pages for each partner.
Answer partners' product questions on demand, and build demo scripts and sample data for a prospect.
Route account overlaps to the right rep, prepare meeting briefs, and sync referrals with cloud partner portals.
Score partner health every week, flag partners turning yellow, and draft quarterly reviews.
Select a stage to see what it involves.