Most early stage AI consulting firms don’t have the budget to hire expensive machine learning talent. For non-technical founders who can’t do the ML engineering themselves, this means getting creative when it comes to AI project delivery.
While there are plenty of non-technical AI services that a consultant can provide – much of the budget for AI services (for better or for worse) is to be allocated to technical problems, such as:
- Building an AI model for X (fraud detection, product recommendation, etc)
- Integrating a vendor solution into existing IT and data infrastructure
- Augmenting an existing data science team to complete a complex project
What we see in the market is that the technically-capable firms are often the ones who get the opportunity to inform and educate enterprise leaders and lift their AI fluency.
This isn’t because technical firms are the only firms capable of this kind of critical AI education. Rather, it’s because enterprise firms think they need technical expertise, and they will often take education only as a side effect, a “bundled” element of a larger AI initiative. There is limited budget for “education,” and certainly for education not connected directly to an existing AI initiative.
So what is a non-technical AI consulting founder to do?
Enter the AI delivery partner model.
The AI Delivery Partner Model
Non-technical AI consulting founders can find small or mid-size machine learning development shops – or a handful of talented freelancer/“moonlighter” ML engineers, and form a mutual agreement that if that consultant can sell work of X type, both parties will work together on the project in tandem.
This avoids the need to hire expensive talent upfront, and allows a consultant to handle technical projects without yet having technical staff. As common as this approach is, it isn’t as easy as finding an ML engineering firm and asking “Hey, if I sell some work could you work on it for a fee?” In the sub-sections below, we’ll explore the steps involved in building a partnership with a technical AI firm.
The remainder of this chapter should read as a set of actions steps to follow in order:
Criteria of a Good AI Delivery Partner
Before building a prospecting list of potential technical delivery partners, it’s important to determine the specific skill and experience mix you’re looking for. A good technical delivery partner:
- Has direct experience with the kinds of projects you plan to win business for. This might involve industry experience (insurance, retail, etc), or experience with specific AI applications or approaches (NLP for conversational interface, satellite and drone computer vision, etc).
- Is small enough to have an interest in a partnership. Larger firms with sizable business development teams and existing sources of lead flow are less likely to want to juggle one-off introductions from a single outside partner. We’ve found that firms under 20 employees tend to be the best fit for this kind of partnership. On occasion, an individual AI freelancer can be approached with this kind of proposition, but these individuals are often unable to commit fully to a project as they are usually already (a) bogged down with projects of their own, or (b) working full time and have limited time for side projects. Smaller AI consultancies are usually a better fit.
How to Find an AI Delivery Partner
Finding small AI consulting firms is as easy as a Google search. Alternatively, many AI platforms (like DataRobot) have a “partners” page that lists the consulting firms that they work with, and this can also serve as a directory of potential partners. In addition, platforms like UpWork (or their alternatives) often have small ML consulting firms bidding for projects – and these consulting firms are almost certainly open to referrals.
As you comb the sea of possible consulting firms, consider:
- Which of these firms has the kind of talent and experience you need?
- How experienced (in academic degrees or industry roles) are the founders of this firm?
Plan to reach out to around 10 potential delivery partners – and plan to speak in depth with at least 3-4 partners before deciding who might be the best initial partner to work with. Personal chemistry and “cultural fit” elements are important here. Speaking with multiple partners will allow you to determine where personalities and incentives are most aligned and with whom your trust feels strongest.
1. Reaching Out to Prospective Partners
When connecting with potential AI delivery partners by email or LinkedIn – include the following elements in your outbound message:
- How you could directly benefit them soon. It’s best to reach delivery partners when you already have a good sense of the market you’ll be going into – or even when you have some warm conversations in the works about AI projects. A partner should be convinced that what you bring to the table (trust, business smarts, and most importantly lead flow) is extremely valuable. They’re only interested in partners with a track record of success who are likely to continue that track record of success with them. Make it clear what kinds of contacts you have, and what kinds of conversation and pipeline you’re already developing.
- A reason you selected them. Nobody likes a blast email – make it clear to a prospective partner why you believed they might be a food fit specifically. Was it their academic background, their experience in a specific industry?
- An ask to connect. Ask directly if the partner would be interested in connecting and possibly forming a partnership.
Most of the usual advice about cold outreach applies here. Subject line selection is important, and being succinct in your wording is also important. These outreach efforts should be seen as a campaign, not a one-time touch. Consider a cadence of 3-4 email and LinkedIn messages over the course of a month in case they miss the first email, or are not yet convinced enough to reply to you.
2. Agree to a Working Arrangement
Once you connect with a handful of prospective partners, decide on the ways that you two will work together. A basic draft of a working agreement needn’t be complex, but should handle the following elements at minimum:
- Responsibilities for sale and delivery. Who brings in the leads? When is the delivery partner involved in the sales process, and when are they not involved? What ongoing project responsibilities will fall in the hands of the consultant, and which in the hands of the delivery partner?
- Split of revenue of profits. This might vary depending on circumstances. For example – if the delivery partner themselves introduces the consultant to a lead to be “closed” – is the revenue split different than if the consultant brings in a lead that the delivery partner would have never found otherwise?
- How to represent your partnership. How will you both refer to your partnership publicly, and to prospects? What is the coherent story you’ll tell about how you work together so that prospects will be able to rest confidence in your “team”, despite you not all being employed by the same, single company.
- What kinds of projects you’ll say “yes” and “no” to. Not all AI projects are a good match for an early consultant and delivery partner team. Projects outside the areas of industry or technical competency could mean poor client results, a poor reputation, and lost margin on early projects. Define a bounding box for how to say “yes” and “no” to projects, so that you can mutually feel confident about which projects you jump in on together.
3. Fielding New Business
Once you have an agreement in place and a shared understanding of your responsibilities, it is usually the consultant’s job to enter the market (often through their own warm contacts) and find new business opportunities.
Discovering AI opportunities and framing the ROI of AI is usually the responsibility of the consultant. Though the delivery partner can be brought in wherever they might add value or engender trust in the process – the consultant’s value in this partnership is sales above all else.
4. Deliver Early Projects and Develop Collaboration Run Books
While early projects should ideally be profitable, they also serve the broader purpose of developing a rhythm of work and collaboration between the two parties.
Reflecting on the sales process and delivery process allows the team to build runbooks for future collaboration – finding a pace and methodology for making future sales and delivery easier. These “retained learnings” should be seen as one of the most important kinds of ROI in early AI projects.
AI Delivery Partners – Perspectives from Emerj Interviews
Essentially all that we’ve learned about the general idea of AI delivery partners has come from direct interviews with bootstrapped AI services leaders. The chicken-and-egg problem (hiring expensive data science talent vs. fielding AI-related projects) exists for all AI companies, but is especially prescient for services companies founded by non-technical leaders.
Below, we’ve listed some of the episodes from our AI Consulting Podcast that include stories of “bootstrapping” AI talent.
Luv Tulsidas of Techolution discusses his strategy for keeping a “bench” of data science talent to be pulled in full time only once work has been booked:
Krunam CEO Chris Wexler discusses the origin of his AI business model, and how he partnered with a talented technical team to start his company: