Document review and beyond: How the latest AI developments can get more from your data

The vol­ume of data that organ­i­sa­tions must deal with in legal dis­putes and com­pli­ance inves­ti­ga­tions con­tin­ues to increase expo­nen­tial­ly year on year, as the sources of data also evolve rapid­ly. With the ana­lyt­i­cal tech­nol­o­gy and meth­ods for review­ing data quick­ly becom­ing out­dat­ed and inef­fi­cient, busi­ness­es with already stretched resources can find it hard to keep up.

Cur­rent­ly, when­ev­er a legal mat­ter or inves­ti­ga­tion aris­es, poten­tial­ly rel­e­vant data is col­lect­ed from dif­fer­ent com­pa­ny sources into a sin­gle data­base, where legal teams attempt to weed out irrel­e­vant data before spend­ing thou­sands of hours review­ing doc­u­ments one by one, cat­e­goris­ing and flag­ging impor­tant or sen­si­tive data. When the next case aris­es, the whole process is repeat­ed, often with the same data col­lect­ed, culled and reviewed again, dupli­cat­ing the cost.

This silo­ing of infor­ma­tion means there is no way of shar­ing data and learn­ings across mat­ters. A lawyer in one case may spend weeks review­ing thou­sands of doc­u­ments to iden­ti­fy those that con­tain pro­tect­ed per­son­al infor­ma­tion of clients, despite those same doc­u­ments hav­ing pre­vi­ous­ly been reviewed and redact­ed for that infor­ma­tion on a dif­fer­ent mat­ter. Such unnec­es­sary work and rework cre­ates incon­sis­ten­cy, plus the poten­tial for sen­si­tive infor­ma­tion to be acci­den­tal­ly pro­duced.

The need to pro­tect sen­si­tive infor­ma­tion is a real con­cern to com­pa­nies with ever-increas­ing data vol­umes. The legal and reg­u­la­to­ry frame­works that pro­tect per­son­al data are rapid­ly chang­ing and becom­ing increas­ing­ly strin­gent, with a grow­ing focus on the Gen­er­al Data Pro­tec­tion Reg­u­la­tion in Europe and, in the Unit­ed States, recent­ly enact­ed Cal­i­forn­ian pri­va­cy laws cre­at­ing sim­i­lar­ly strict reg­u­la­tions that oth­er US states are poised to fol­low.

These reg­u­la­tions bring incred­i­bly high mon­e­tary penal­ties if com­pa­nies inad­ver­tent­ly dis­close per­son­al infor­ma­tion and yet, when deal­ing with unstruc­tured data, it can be dif­fi­cult to com­ply.

“It can be hard to pre­vent sen­si­tive infor­ma­tion from being pulled in because organ­i­sa­tions do not know it is there,” says Karl Soby­lak, senior prod­uct man­ag­er at Light­house, which has been inno­vat­ing in enter­prise data for com­pli­ance and legal for 25 years. “For exam­ple, an employ­ee emails a co-work­er and includes a spread­sheet of birth dates, that spread­sheet then gets into the email sys­tem and it is hard to keep it from show­ing up in data­bas­es.”

That spread­sheet con­tain­ing per­son­al data can get pulled into mul­ti­ple review data­bas­es for dif­fer­ent cas­es, increas­ing the risk it gets missed and ille­gal­ly trans­mit­ted to an oppos­ing par­ty.

Old­er gen­er­a­tions of arti­fi­cial intel­li­gence (AI) and ana­lyt­i­cal tech­nol­o­gy are ill-equipped to respond. Most court­rooms and legal pro­fes­sion­als are famil­iar with tech­nol­o­gy-assist­ed review (TAR) but, as Soby­lak explains: “The lim­i­ta­tion of most TAR is it can only be used in the con­text of a sin­gle case and requires sig­nif­i­cant input and turn­ing of data that can take sev­er­al weeks. Then, when you’re done with one case, you move onto the next and rein­vent the wheel. So far, we’ve seen that 42% of doc­u­ments are reviewed on mul­ti­ple mat­ters. At the same time, 94% of doc­u­ments being flagged as poten­tial­ly priv­i­leged are not actu­al­ly priv­i­leged.”

Now, advances in AI and machine-learn­ing can help, increas­ing­ly mov­ing on from trou­bleshoot­ing to adding real val­ue.

Soby­lak says more advanced AI can solve the siloed data prob­lem where com­pa­nies pay for data to be reviewed and re-reviewed. “TAR is all based around a sin­gle algo­rithm,” he says. “This new gen­er­a­tion of big data tech­nol­o­gy analy­ses text, meta­da­ta and pri­or lawyer deci­sions using a num­ber of algo­rithms to give con­text around who is talk­ing to who, what their com­mu­ni­ca­tions were like pre­vi­ous­ly and how pri­or review­ers inter­pret­ed that infor­ma­tion, cre­at­ing the abil­i­ty to do a much bet­ter job.”

These advances allow a more holis­tic approach, aggre­gat­ing data from hun­dreds of past mat­ters and using it to spot trends and sup­port data-dri­ven deci­sion-mak­ing. Soby­lak says: “Not only does this speed up doc­u­ment review and deliv­er cost-sav­ings, it allows for con­sis­ten­cy with­in and across mat­ters, which can sig­nif­i­cant­ly low­er a company’s risk.” 

This can also be invalu­able at low­er­ing the risks asso­ci­at­ed with sen­si­tive infor­ma­tion. Because it can lever­age mul­ti­ple algo­rithms across pre­vi­ous lawyer work prod­uct, this new AI can more accu­rate­ly find the types of per­son­al infor­ma­tion that need to be redact­ed or with­held. If a doc­u­ment was flagged as con­tain­ing per­son­al infor­ma­tion on one mat­ter, it will be flagged on sub­se­quent ones.

It can also help iden­ti­fy data pro­tect­ed by legal pro­fes­sion­al priv­i­lege, which can require more than the basic iden­ti­fi­ca­tion of key­words and calls for addi­tion­al con­text. New­er AI tech­nol­o­gy is capa­ble of iden­ti­fy­ing the con­text in which a con­ver­sa­tion between two lawyers takes place and whether it is legal­ly priv­i­leged, lever­ag­ing knowl­edge learnt from deci­sions made on pre­vi­ous mat­ters.

There is the poten­tial to solve many of the big data chal­lenges and dras­ti­cal­ly improve cost effi­cien­cy for com­pa­nies. Should this wor­ry lawyers, who were pre­vi­ous­ly paid to con­duct all those hours of reviews and re-reviews?

Rob Hellewell, vice pres­i­dent of Light­house and a for­mer prac­tis­ing lawyer at a large inter­na­tion­al law firm, says: “AI is not going to replace lawyers. This tech­nol­o­gy is about enhanc­ing lawyers’ abil­i­ty to sep­a­rate the chaff and get to the wheat. Our focus is on aug­ment­ing their abil­i­ty to find rel­e­vant, priv­i­leged and sen­si­tive infor­ma­tion faster, con­sis­tent­ly and at less cost.”

Hellewell argues that lawyers often fail to appre­ci­ate the data aspects of their roles, but jump on board once they see results. “Imag­ine there’s a new mat­ter with a mil­lion doc­u­ments to review and you can start by elim­i­nat­ing the need to review ones that have been looked at pre­vi­ous­ly,” he says. “Lawyers can see those ben­e­fits imme­di­ate­ly.”

In addi­tion to sim­ply mak­ing day-to-day tasks eas­i­er and work prod­uct more con­sis­tent, new­er AI solu­tions con­tin­ue to break new ground and add more intrin­sic val­ue to legal prac­tice.

“Nor­mal­ly we are just pro­vid­ing the infor­ma­tion that we are required to pro­vide to set­tle the case. But AI now gives us the oppor­tu­ni­ty to look at how we might have pre­vent­ed a dis­pute from aris­ing or iden­ti­fied a prob­lem soon­er,” says Hellewell.

From a com­pa­ny-wide stand­point, the lat­est AI offers poten­tial to reduce risk and improve cost effi­cien­cy, while from a legal team per­spec­tive, it can unlock valu­able legal insights from com­pa­ny data and enhance the crit­i­cal con­tri­bu­tion of in-house lawyers and exter­nal law firms.

For more infor­ma­tion please con­tact info_london@lighthouseglobal.com

Pro­mot­ed by Light­house