The Impact of AI on SPM

Understand the advantage of artificial intelligence and how to enable SPM owners and revenue leaders to determine the most effective ways for their businesses to embrace the impact of AI on SPM.

Artificial Intelligence (AI) is changing our world. Though there have been many technology hypes in the last few decades, think what3words, blockchain, etc. AI and Large Language Models (LLMs) are different. Unlike these previous hypes, LLMs are already effectively addressing real world problems,  from writing parking complaints to the council to accelerating coders (those writing SQL). This moment is perhaps reminiscent to the early days of the internet, where we are  aware of the power, but are still determining how to use it effectively.

AI will undoubtedly change how we interact with the world around us. Businesses are rushing to change existing processes and the field of Sales Performance Management (SPM) is no exception to this. Discerning the hype from the true potential impact, however, is bewildering especially due to the abundance of terminology in the space.

This paper written in October 2023, aims to clarify the role of AI in Sales Performance Management (SPM). It provides preliminary predictions to enable SPM owners and revenue leaders to determine the most effective ways for their businesses to embrace this transformative technology.

Want to disscuss the impact of AI on SPM with your peers and our team of experts? If so, let us know and we’ll be in touch with event details soon!

What is AI?

The key types of AI and terminology that we will use in this paper are described below:

Artificial Intelligence (AI): This refers to the range of abilities that computers are developing to mimic human-intelligence. Examples include problem solving, forecasting, playing chess and more recently, responding to questions.

Machine learning (ML): ML is a subset of AI where explicit programming isn’t necessary. Instead of telling the computer the rules to identify, say, a cat (like having two eyes, whiskers, and a tail), it learns these rules from input data.

Large Language Models (LLMs): LLMs are a specific subset of Machine Learning. They learn from vast amounts of text data to predict sensible responses to questions.

Predictive Analytics (PA): This is a capability of artificial intelligence that may rely on machine learning. This portion of AI is the capability of predicting events based on historic data. This is typically dependent on large bodies of data, so called Big Data.


While many leading SPM technologies incorporate AI features, the adoption has been slow. We believe this slow adoption has been due to the focus on predictive analytics. We believe that adoption will increase as the technologies incorporate the benefits from the breakthroughs in LLMs.

Why has uptake of predictive analytics been slow? It has been surprising to many that the uptake of predictive analytics has been slow in SPM. It would seem given clean data (people are paid on it), important questions to answer (what compensation plans work, at what cost), and significant potential upside, uptake should have been ubiquitous and rapid. Why has this not been the case? We believe this lag is due to several reasons.

• Time to demonstrate benefit – By its nature, predictive analytics requires a prediction to come true for benefit to be validated.

• What to investigate – Most organizations are not used to interrogating their sales comp data (often their most accurate data), and do not set aside time/effort to try and answer these questions. Insightful questions could be: What is the optimal commission %? What is related to sales rep churn? What employees are at risk?

• Right Data – In time – Though the data that SPM systems hold is often of high quality, this often does not include the latest or complete data. Additional data will enable greater exploratory and explanatory power, and hence benefit. Examples of additional data that could unlock further insights are forecast data, and supporting data such as sellers, products, and customers.

• Exploration not enough – Exploring and showing correlations has some benefit, such as identifying at-risk salespeople but this may not be enough benefit for the time investment.

• Experimentation not allowed – To prove causality not just correlation experiments are needed. Organizations are hesitant at doing this as they want to be optimal for everyone, even though the best for their business may differ from general best practice.

• Experimenters Required – Running useful experiments is hard and requires a different skill set and focus which many SPM owners do not have. As the role of data analyst becomes more ubiquitous, and the technological barriers reduce, expect this challenge to be reduced.

Due to the potential of predictive analytics, we believe that uptake of this use-case will increase as the technological barriers reduce, and data analysis becomes more of a cultural norm.


The LLM breakthrough, which became widely apparent in 2023, has been the progress of the large language models such as ChatGPT and Bard. Large Language Models can respond to complex user questions and perform complex tasks such as achieving 90th percentile results in the bar exam or, looking at a more frivolous example, can pass sommelier exams.

This paper will handle benefits from the benefits of predictive analytics and LLM.


In the short term, AI in SPM will be focused on predictive analytics and basic chatbots. In the longer term, we expect increased automation, improved plan  communication and beneficial coaching advice.

Here are our expected use cases, organized by their anticipated rollout timeline:

1. Identifying Data Anomalies – This involves spotting anomalies in both inbound and outbound data, particularly in payout results. This will be using predictive analytics rather than the LLMS. This will work by creating expected distributions of values, and highlighting anomalies based on this.

2. Assisting with Sales Plan Effectiveness Analysis – This includes identifying sales team behaviors that impact sales using predictive analytics rather than the LLMS. Easier to use predictive analytics, and improved data integration will lower the barrier of entry for this analysis. Though the correlations will be useful in their own right, organizational willingness to experiment and careful plan design will be required to show causality and really reap the full benefits.

3. Enhancing Sales Forecasting Accuracy – Leveraging predictive analytics AI will increase the accuracy of both the sales forecasts and the expected compensation spend. This prediction of cost will enable decisions to be made on plan parameters to be set with better knowledge of distribution of expected costs. This will reduce overspend.

4. Assisting with Quota and Territory Analysis – In territory setting, there can be considerable optimization problems, for example, if the cost/time to visit a client is dependent on the distance to the client, what geographical territories are optimal for the maximal revenue, and what quota is reasonable for each quota. This problem can be solved by some of the techniques behind predictive analytics.

5. Developer Chatbots – Trawling through developer documentation will be a thing of the past, instead of being directed to an article listing formulas, once you enter “what formula do I use to calculate an attainment?”, the chatbot will answer directly. This is relatively simple to develop as the documentation for a solution can be fed into the Large Language Model as part of the prompt. Try out “how do I get the end of month from a date in SQL” in ChatGPT if you want to see this already working. We anticipate that design documentation may also be fed into these to allow admins to better under existing code and dependencies. An example prompt: “What technical considerations are there if I make the plans monthly”, may be useful.

6. Sales Rep Inquiry Handling – Chatbots will assist in handling common commission inquiries and guide individuals through help pages. AI will increase the  accuracy and proportion of inquiry responses that can be automated. This will save both sales rep and admin time, though we believe that the human touch will be required for more complicated inquiries.

7. Plan Documentation Creation – Perhaps the easiest in fixed data model tools, plan documents will be created on the rules configured, and if required, personalized summaries of the plan for the sales reps. This removes the risk of disputes that occur when plan documents and rules are misaligned.

8. Plan Documentation – Confirm Understanding not Acceptance – One of the key factors that drives the effectiveness of a plan, is the understanding of it. Until now, this confirmation of understanding has been tricky to confirm, sales reps often accept plans without understanding them. We see an interesting use case here, where end users could ask questions about their plan before accepting it of an AI agent. An interesting, if perhaps slightly out-there idea, would be
using AI to create customized tests for users to validate understanding before sign-off, similar to the way that the IT compliance quizzes are run.

9. Sales Representative Enablement – Sales representatives can get useful answers through advanced chatbots, which deals have been neglected, or which may be at risk, or what deals are needed to close for a specific payout.

10. Operational Automation – Admins will be able to complete operational tasks quicker through text prompts for general operation tasks. For example, the command “run calculations from April to June 2023, send me an email when done with a link to the report for John Smith”. It is expected that instructions for changes configuration will require an intermediary approval step.

11. Configuration Automation – In the longer term, admins will be able to complete configuration tasks through text prompts. For example, “create a Business Development Representative (BDR) plan which works the same as the Inside sales plan, but with no revenue component.”

12. Coaching Assistance – Managers can request more complicated analysis, imagine a prompt such as, “which deals should I prioritize”, “what can I do to improve Abigail’s sales”. This will be a combination of LLM and predictive analytics.


With any new technology there are new risks. Although some of these are substantial, due to the magnitude of the expected benefits, we expect these to be resolved or mitigated in short order.

1. Legal Concerns – Most AI LLMS are offered by third-party providers hosted in the United States. Incorporating this technology into any tool may raise legal data sharing issues. The data stored in SPM systems is highly sensitive, and it’s crucial to have the right legal agreements and deployment options in place. Companies would not want their data leaking and being publicly available. To mitigate this, we must carefully consider the data transmitted to the large language models. We expect that AI agents may be deployed semi-privately. There is a tension here, AI providers will want the right to access the data in the tools to improve AI performance (refer to the recent controversy of Zoom Terms of Service for an example).

2. Overreliance Risk – As AI becomes increasingly capable, there’s a risk of overreliance. Once AI reaches a point where it’s as good as or slightly better than humans, we may come to depend on it. Think about how we rely on GPS for navigation. We foresee scenarios where AI generates drafts, and it’s the user’s responsibility to validate them. This will be especially important for actions that are difficult to reverse (e.g., running a process). Without proper processes in place, users might click through wizards without realizing the potential consequences. Another overreliance problem may be design integrity, if AI can build short term fixes easily this may lead to design principles being ignored.

3. High-Profile Mistakes – Though the likelihood of mistakes will ultimately be reduced, the root cause analysis attributing failings to AI will be unsatisfying to many. Testing phase of a fully AI built system (a while of yet) would also be impacted, without understanding how AI works. It becomes tricky to test fully. Perhaps think of the widely reported 2016 story where a car mistook a truck for the sky. When introducing AI, as with any change, careful management is essential to ensure that responsibilities are clearly defined to avoid potentially costly mistakes.

4. Bias and Unpredictability – AIs respond, but their reasoning can be obscure. This is largely based on the data they were trained on and their training process. It is challenging to guarantee that we will be satisfied with their responses. We can see how this could be uncomfortable if it starts treating different personas differently. To mitigate this, we must carefully consider the data transmitted to the large language models.


We expect that within the next two years, most companies will integrate AI into their SPM systems. Which SPM technologies will develop this soonest?

The SPM systems that will develop AI capability soonest will be:

1. Investing Strategically –
They will be actively investing significantly in their product roadmap to incorporate AI features.

2. High Potential for Gain – These technologies stand to benefit significantly from AI integration. This could include new entrants seeking a competitive edge or established players with a substantial client base, enabling them to offer premium AI-enhanced services or enhance customer satisfaction.

3. Easily Integrated with AI – Being the easiest for AI to learn, and hence the lowest cost and complexity. Technologies featuring structured data models and minimal configuration complexity we expect to be the easiest to adapt and integrate with AI.

Several factors that could influence ease of integration:

Training Data Availability – Training data, perhaps from knowledge bases and community forums, but potentially also created examples. We may see contractual changes here, to allow technology providers the use of customer data for training purposes, with some new frontiers opening in data protection language. Refer to the recent controversy of Zoom Terms of Service for more context.

Transference Ease – How readily the AI can apply its existing knowledge to the tool. For example, the AI may have ability in SQL that it can reapply easily.

Complexity Considerations – Complexity of the use case and complexity of the tool will both have an impact here. Basic use cases could be implemented without additional training of the LLM. In terms of technology, simpler tools may have a higher knowledge requirement (more formulas to remember) but lower conceptual complexity (e.g., fewer tables). Conversely more free-form tools may require less prior knowledge but necessitate a deeper understanding of concepts.


Five years ago, the battleground for SPM vendors revolved around the ability to calculate complex incentives and offer flexible, visually appealing reports. AI is poised to usher in the next competitive wave in the SPM space, and it will be fascinating to witness how SPM vendors differentiate themselves here.

In addition to the new functionalities mentioned earlier, we expect AI to reshape the SPM technology landscape in the following ways:

• Separated License Costs –
As AI technology matures, companies with large existing customer bases may sell AI as an extra module at additional cost. Potentially a significant benefit of being an early adopter here.

• Direct Data Integration Imperative – Not all AI analysis will be done within the SPM tool, some will accumulate to the system housing most data, such as the CRM, ERP, and marketing automation platforms. This will drive vendors to develop more robust APIs and integration capabilities, leading to a more interconnected technology ecosystem. SPM technologies with weaker data integration capabilities will adapt or see difficulties.

• Broadening of SPM Offerings – As AI becomes more powerful, there will be a benefit of using it in areas connected to SPM. Providers that offer territory management and sales planning capabilities will ultimately be able to offer more powerful AI features.

• Humans as Validators – Current AI usage often requires human validation. The speed and extent to which we remove humans from the loop, and in which cases, remain open questions. In the midterm, we anticipate humans spending more time confirming actions required by AI, further enhancing its performance.

• Shorter Implementation Timelines – Utilizing design tools and code generation will shorten the window and eventually cost of implementations. Validations will still be key and ensuring that model and design principles are adhered to. Quicker deployment and changes may allow more interaction and feedback with business users on changes and their impacts.

• Shorter Feedback Loops – Explore, Experiment, Exploit loop will be shortened. This will lead to a greater understanding of causes of less profitable selling leading to better plans and better selling.

• Code Language Evolution – In the midterm, expect that SPM technologies may move to more general technologies for customizations and move away from custom dialects such as xSQL and Groovy, to take benefit of more advanced AIs trained on more standard code dialects.

• Enhanced Reliability Back-end teams managing cloud architecture will also benefit from AI. In the midterm, we expect self-healing tools further reducing down-time.

• Redefined Sales Roles – As AI takes over more administrative and data-driven tasks, the role of sales reps will evolve. Sales teams will devote less time on routine tasks and more time on building relationships, understanding customer needs, and providing value-added services.


OpenSymmetry anticipates that the impact of AI on implementing an SPM technology will be significant in the short, mid, and long term.

1. Reducing Build Time and Costs

In the short term, the cost of certain customizations will drastically reduce. New customizations made through widely used coding languages such as SQL or JavaScript will benefit the most. Where coding languages are specific to the SPM technology, less cost reduction expected.

In the mid to long term, configuration costs will decrease as administrative text prompts become more robust. An example text prompt illustrates the potential: “Create a direct and indirect credit rule that filters for new business transactions”. An intermediary approval step will be required here. There is a risk here that moving too fast may break too many things. See Overreliance Risk on Page 4.

2. Reducing Testing Time and Costs

In the short term, expect that testing timelines will reduce slightly. It will become common practice for test data to be created using AI instead of Excel spreadsheets. Currently this can be done readily for small data sets using ChatGPT for larger data sets this is trickier.

In the midterm, administrative operational text bots will further reduce the cost of executing test cases.

3. Reducing Expectations

In the longer term, As AI handles simpler tasks or accelerates their completion, only those with extensive knowledge of technologies or deep consulting expertise will continue to provide value. The expectations of consulting resources will increase.


AI capabilities are poised to deliver significant benefits to users of SPM technology. Some of the key advantages we anticipate include:

1. Sales Time Savings – Sales people will be able to understand their compensation quicker, through the use of AI chatbots, further reducing shadow accounting

2. Improved Decision Making through Accurate Forecasting – Predictive analytics will improve forecasts, crucial for executive decisions and expense associated with commission accrual

3. Improved Coaching – Ability to coach sales reps will improve with managers and sales reps receiving custom coaching advice

4. Improved Plan Understanding – Chatbots will enhance plan communication, reps will under stand their plans better leading to better alignment to sales plan

5. Less Incorrect Payouts – There will be fewer compensation errors through anomaly detection, leading to less shadow accounting and a more motivated sales force

6. Fewer, more powerful administrators – Inquiries will be handled quicker and more efficiently by chatbots avoiding potential disputes

7. Tailored Compensation Plans – Although a tension will remain between plan complexity and understandability, certain comp plans which are not used due to perceived maintenance, communication difficulty or implementation overhead will be used. Think of monthly plans as an example

8. Reduced Implementation Costs – Implementation costs for the same scope will reduce

9. Boosted Profitability – Over time compensation plans will improve based on demonstrable effectiveness, ultimately driving more profitable revenue for businesses

Want to disscuss the impact of AI on SPM with your peers and our team of experts? If so, let us know and we’ll be in touch with event details soon!

pdf on computer icon

Want The PDF Version?

OpenSymmetry Global Offices

About Us

OpenSymmetry enables clients to achieve greater operational efficiency and get better sales results. We are a global consulting company specializing in the planning, implementation, and optimization of industry leading technology suppliers of sales performance management solutions.