The most common question we get when describing August's clause deviation detection is: "Deviating from what exactly?" It is a good question. The answer tells you a lot about why we think this approach to contract review is more useful than the alternatives.
The benchmark for a deviation flag in August is not a generic legal standard. It is not the American Bar Association's model form, not the NVCA model documents, not whatever language is most common across contracts available in public databases. The benchmark is how your firm has written that clause in comparable past deals.
The Baseline Problem in Generic Review Tools
Most AI-assisted contract review tools work by comparing a draft against either a set of curated playbook provisions or a model that has internalized what "standard" language looks like across a broad training corpus. When they flag something, they are flagging a departure from that external standard.
There are situations where that is exactly what you want. If you are asking whether a clause is legally unusual, whether it departs from common practice in a given jurisdiction, that external comparison can be informative. But in transactional work, the more pressing question is usually different: does this clause depart from what we would normally agree to? That question requires a different benchmark.
A clause that is perfectly standard in the abstract might still be a material concession for a particular firm, if that firm has historically held a tighter position. Conversely, a clause that looks unusual by generic standards might be entirely consistent with how a firm has handled a specific transaction type. The comparison to generic practice does not answer the question that actually matters during negotiation.
How August Determines a Comparable Deal Set
When a new agreement comes into August for review, the system identifies the relevant comparison set from your precedent bank by examining several factors simultaneously: the transaction type and deal structure, the relative positions of the parties, the industry sector, the governing law, and the economic parameters of the deal where those are discernible. It is not a simple keyword match. The system builds a representation of the current deal's characteristics and identifies the past deals in your bank that most closely resemble it across those dimensions.
For a technology licensing agreement where the firm is representing the licensor, the comparable deals might be the twenty-three technology licensing matters the firm has closed over the past four years where it represented licensor-side clients. For a commercial real estate lease where the firm represents a tenant, the comparables are tenant-side commercial leases in similar markets. The composition of that comparable set shapes every deviation flag that follows.
This matters because what counts as a deviation varies significantly by transaction type and party position. A damages cap at two times annual fees is a deviation in one context and well within normal range in another. August's flags reflect your firm's actual experience in the specific deal context, not a single cross-context average.
What a Deviation Flag Contains
When August flags a clause, the flag includes three pieces of information. First, the specific dimension on which the clause departs from your bank's comparable deals, stated precisely: "Cap is set at 3x annual fees; your firm used 2x or 1.5x in 7 of 9 comparable deals." Second, the citation to the past deals that inform that assessment, which the reviewing attorney can open and examine. Third, the range of variation that appeared in those comparable deals, which shows whether your firm's position on this clause has been consistent or has varied with circumstances.
That third piece is sometimes the most useful. If your firm has used a range of caps across comparable deals, the flag does not tell you this clause is wrong. It tells you this clause lands outside the range of what you have accepted before and prompts the attorney to consider whether the circumstances of this deal justify it. If your firm's position on this clause has been very consistent, a departure is a stronger signal that something needs attention.
The Difference Between a Flag and an Objection
We want to be direct about something. Clause deviation detection is a review tool, not an approval system. August flags clauses that depart from your firm's historical practice. It does not tell you whether to accept or reject those clauses. That judgment belongs to the attorney who understands the full negotiating context, the client's priorities, and the deal-specific constraints that may make a particular concession appropriate.
In practice, this means August will flag some things that the attorney will decide are fine given the circumstances of this deal. That is expected and correct. The value is not in eliminating those flags. The value is in ensuring that the attorney makes a conscious decision about each departure rather than missing it during a fast review pass, and in ensuring that when a departure is accepted, the record shows why it was accepted rather than simply showing the final agreed language with no context.
False Positives and the Calibration Challenge
Any deviation detection system will produce false positives in some proportion, and we are honest about that. The most common source of false positives in August is when the comparable deal set is too small or too heterogeneous to produce a reliable baseline. If your firm has only closed three deals in a specific transaction type and governing law combination, the baseline for that category is not robust. August will still flag deviations, but the attorney should weight those flags accordingly.
We have also found that practices which undergo significant evolution in their standard positions over time, shifting from one preferred cap structure to another over several years, can produce misleading comparisons if the older deals dominate the bank by volume. August lets you filter the precedent bank by date range, so a firm can choose to compare against more recent practice if that better reflects current standards.
Where This Is Most Useful in Practice
From what we have seen building August alongside transactional practices, clause deviation detection adds the most value in three specific situations. The first is high-volume review, where an associate is moving through a stack of similar agreements and the cognitive load of holding the firm's standard position in mind across all of them is genuinely high. A consistent external check against the bank reduces the error rate in those conditions.
The second situation is when less experienced attorneys handle matters that are within a type the firm does regularly but outside the specific partner's practice history. The associate might know the general framework but not the specific positions this firm has taken in this deal type. The bank serves as institutional memory that would otherwise require interrupting a senior attorney for guidance.
The third situation is the counterparty-proposed redline. When a counterparty proposes changes to multiple clauses simultaneously, the sheer volume of changes can make it difficult to assess which represent material departures versus minor language preferences. Routing that package through August against the comparable deal set gives the reviewing attorney a prioritized view of which changes represent the most significant departures from how this firm has previously handled the same issues.